Datasets:

rulins

/

pes2o_v3

like 0

in the language network during incremental speech comprehension, largely due to the lack of evidence for characterizing the spatiotemporal dynamics of neural activity. Further research is needed to understand the detailed neural mechanisms underpinning these important effects. Supplementary Material Supplementary material can be found at Cerebral Cortex online. Notes We thank Dr Barry Devereux for his help in the early stages of this research.

221404245

Analysis of community participation at Community Forestry Group (HKm) in Forest Management Unit (KPH) Region XIV Sidikalang, North Sumatera Province The community participation is essential for the success of forest management in the Forest Management Unit. This study aimed to identify the form and level of community participation in supporting the Community Forestry Program. This research was conducted from June to September 2018 by using survey and interview methods. The data collection used by reviewing documents, observation, questionnaires, and interviews. The results of this study revealed that the highest participation form was ‘engaging in social activities’ chosen by 91% of respondents in Aor Nakan Village, 91% of respondents in Kuta Tinggi Village, and 100% of respondents in Sibongkaras Villages. Introduction The Minister of Environment and Forestry Regulation No. 83 of 2016 confirms that Social Forestry is "a system of sustainable forest management carried out in state forest areas or customary forests/customary forests implemented by local communities or conventional law communities as the main actors to improve their welfare, environmental balance and social and cultural dynamics in the form of 1 ) Village Forest, 2) Community Forest, 3) Community Plantation Forest, 4) Community Forest, 5) Customary Forest and Forestry Partnership. Through this policy, there are several things that the government wants to achieve, namely: (a) creating and accelerating equitable access and distribution of forest resource assets; (b) resolving tenure conflicts in forest areas; and (c) reduce poverty and improve the welfare of people living in and around forest areas. The emergence of a sustainable development paradigm indicates

218804364

the existence of two perspectives. First, the involvement of local communities in the selection, design, planning and implementation of programs or projects that will color their lives, so that can be guaranteed that the perception of local communities, attitudes and mindset patterns and values and knowledge fully considered, while the second is making feedback which is essentially an inseparable part of development activities [1]. The concept of sustainable development which involves community participation in the development process has the goal of improving the welfare of rural populations, including those living in and around forest areas, participating in decision-making and being able to 2 change their livelihoods. The participatory and pro-poor approach in rural communities is known as the populist approach [2]. The community participation is absolutely essential for the success of a development program. It can be said that without community participation any development will be less successful. It was further explained that the community participating in development activities would go through a learning process [3]. Therefore, communities need to experience a learning process to find out opportunities to participate in the development process, and often their abilities and skills still need to be improved in order to take advantage of these opportunities. In 2017, the Minister of Forestry and Environment has issued a Community Forest Utilization Business Permit (IUPHKm) to 3 (three) Forest Farmer Groups (KTH) in Pakpak Bharat District. So far there have been no studies or studies that have analysed the factors that support community participation and the form and level of community

218804364

participation in this program. Therefore, it is crucial to conduct a study on "Analysis of Community Participation in Community Forest Program (HKm) in the Forest Management Unit of Region XIV -Sidikalang, North Sumatra". The objective of this study is to identify the form and level of community participation in supporting the HKm program. Time and place This research was conducted in Aor Nakan Village, Kuta Tinggi and Sibongkaras, buffering Sikulaping Protection Forest (Reg. 71), Pakpak Bharat District, which had obtained an IUPHKm decree from the Minister of Environment and Forestry. This area is part of the KPH Region XIV Sidikalang -North Sumatra. The primary and secondary data were collected from June to September 2018. Research method This research was used to survey and interview methods. Data types and source a) Primary data, consisting of actual condition data at the study site such as forest cover, forest use patterns, community socioeconomic conditions, employment, income, education, land area, and others. b) Secondary data, the supporting primary data, consisting of the general condition of the study site, the long-term management plan of the KPH Reg. XIV, area maps, topography, soil types, and other supporting data. Data collection technique Data collection techniques are carried out through document review, field observations, questionnaires, and interviews. a) Document review. b) Field observations. c) Questionnaire, based on a Likert scale. Each question and/or statement provides several alternative answers to be chosen by respondents in accordance with their perceptions, feelings, and activities. The alternative solutions are processed into quantitative data (given a score). The data

218804364

obtained from the study of documents, observations, and interviews were analysed descriptively qualitatively. While the results of the questionnaire data were analysed in quantitative descriptive using tools such as tables, graphs, and diagrams to explain the level of participation and forms of community participation. Data analysis method Data analysis was carried out to obtain information that could later be used to answer this research question regarding the form, level, and factors that influence community participation in the HKm program. The analysis data are as follows: 1. Forms of participation analysed as support: a) Assistance of Thoughts, which are given by participants in a formal event, meeting or meeting; b) Participation of Personnel, given by participants in various activities for the improvement or development of villages, assistance for others, and so on; c) Participation in Property, which is provided by people in various activities for the progression or development of communities, support for other people who usually in the form of money, food and so on; d) Participation in Skills and Finesse, which are given by people to encourage a variety of forms of business and industry; e) Social Participation, which is provided by people as a sign of communion [4] 2. The level of participation, in this case, is limited to the process of filing a business permit for utilization of HKm (IUPHKm), measured using indicators of involvement in activities and grading using the Likert scale (1-3) namely : a) Participation in planning b) Participation in the implementation of activities c) Participation in monitoring and evaluation

218804364

d) Participation in the utilization of results [5] Each of these indicators will be reduced to four questions, in which each question has three alternative answers represented by values of 1 (one), 2 (two), and 3 (three). Values that describe the "level of participation" of each respondent in each indicator are in the numbers 4 and 12. The evaluation is stated as follows: -High participation, if the number of indicator values is more than 9; -Medium participation, if the number of indicator values is between 5 -8; and -Participation is low if the number of indicator values is less than 4; Participation form The community participation is a process in which people are involved in every stage of the situation that affects their lives [6]. The results of interviews with the community during this study revealed that generally, the terminology of participation was interpreted by the community as a form of community involvement in the Community Forest program, especially in activities directly related to commodity management, because the implementation of the HKm program in this area was facilitated by NGOs. Pesona Tropis Alam Indonesia Foundation (PETAI), where one form of activity is assistance in the cultivation of coffee plants. As for the regional government, in this case, the Forest Management Unit (KPH) Region XIV and the Pakpak Bharat District scope agencies, so far, only involved the community in activities such as counseling by making the community an object or beneficiary without having been involved in planning and decision-making. The results of data analysis revealed that the

218804364

forms of community participation in the three villages were quite diverse, but the highest form of participation was engaging in social activities and was followed by giving ideas/proposals in each meeting. The lowest form of participation is donating assets (money, food, etc.), and this is evenly distributed in all villages ( Table 2). In Aor Nakan Village, out of 45 respondents, 41 respondents or 91% chose to be involved in social activities as the highest form of participation, and seven people or 7% chose to contribute property as the lowest form of participation. In Kuta Tinggi Village, out of 27 respondents, 41 respondents or 91% chose to be involved in social activities as the highest form of participation, and three people or 11% chose to contribute property as the lowest form of participation. Whereas in Sibongakaras village, out of 25 respondents, all or 100% chose to be involved in social activities as the highest form of participation, and no one chose to contribute property as the lowest form of participation. Forms of social activities in the target villages of this study include urup-urup (helping each other in the growing season), planting rice, weddings, planning customary activities, etc. Almost all villagers usually take part in these activities; at least one representative from each family will participate. This confirms that engaging in social activities is the highest form of participation in the region. If related to HKm activities, one of the activities carried out is the improvement of coffee crop cultivation patterns, where there is some new knowledge

218804364

that needs to be practiced, for example, making MoL and making liquid organic fertilizer. The activity was carried out in mutual cooperation in the fields of each member of the HKm group in turn. Furthermore, the form of participation in giving ideas/proposals in each meeting was also high in these three villages. From the interviews conducted, it was found that in every meeting facilitated by the PETAI NGO in relation to the HKm program, the presence of the community was quite high, they were actively involved in giving proposals and even giving critical questions on material or topics they still did not understand. Furthermore, they are also actively involved in the preparation of activity plans, although often in decision making, they submit it to the group leader and/or village head and community leaders who are considered to know better. The existence of PETAI NGOs has a positive influence on community understanding of the form of community participation because of the assistance that is carried out by placing live staff in the village. PETAI NGOs conduct weekly meetings and provide some training related to institutional and organizational management. At each meeting, PETAI staff made it possible and even encouraged the community to express their ideas and ideas. Participation level The community participation in the Community Forestry program in Pakpak Bharat is divided into four stages; planning, implementing activities, monitoring and evaluating and utilizing the results. The planning consists of socialization activities, the formation of institutions (Forest Farmer Groups), preparation of work plans, and the management of HKm

218804364

permits to the Ministry of Forestry and Environment. The Implementation of activities consists of forest utilization activities, preparation of HKm work plans, determination of activity priorities, and implementation of activities. The Monitoring and Evaluation Program consists of monitoring and evaluation activities, conduct regular meetings that discuss the development of activities, and evaluation of activities. The utilization of Results consists of boundaries clarity. Participation level in Aor Nakan Village In general, the level of community participation in Aor Nakan Village is quite high. In the planning and implementation stages, all respondents or 100% with a high participation rate of 100%. At the monitoring and evaluation stage, there were 11 respondents or 24% with medium participation rates and 34 respondents or 76% with high participation rates. Whereas, at the utilization stage of the results, there were 15 respondents or 33% with medium participation rates and 30 respondents or 67% with high participation rates. Detailed results of the questionnaire analysis are presented in Table 3. Level of participation in Kuta Tinggi Village The level of community participation in Kuta Tinggi Village is also quite high. In the planning and implementation stages, all respondents or 100% with a high participation rate of 100%. At the monitoring and evaluation stage, there were 1 respondent or 4% with a low participation rate, 5 respondents or 19% with medium participation rate, and 21 respondents or 78% with a high participation rate. Whereas, at the utilization stage of the results, there were 14 respondents or 52% with medium participation rates and 13 respondents or

218804364

48% with high participation rates. Level of participation in sibongkaras village The level of community participation in Sibongkaras Village is quite diverse. At the planning stage, four respondents or 16% with a 'moderate' level of participation and 21 respondents or 84% with a 'high' level of participation. At the implementation stage, there were 1 respondent or 4% with 'low' and 'medium' participation rates, and 23 respondents or 92% with 'high' participation rates. Furthermore, at the monitoring and evaluation stage, there were seven respondents or 28% with a 'moderate' level of participation. Whereas in the results utilization stage, there were eight respondents or 32% with medium' participation rates, and 17 respondents or 68% with 'high' participation rates. Based on Table 5, the level of community participation is 'high' in planning, implementation of activities, monitoring, and evaluation, as well as the utilization of results. The community participation is needed, as agree with a previous study [7,8]. The community participation in the management of HKm was realized because of the role of PETAI NGOs who assisted by practicing decision making forms. Therefore, in the end, the form and level participation were apparently absorbed and are implemented by the community member. The HKm program in three villages in Pakpak Bharat Regency is a participatory program in the mean of involves the community as HKm manager in various stages, namely planning, implementing activities, monitoring, and evaluating and utilizing the results ( Table 5). The community was involved from the beginning, even in choosing the social forestry scheme to be implemented, which

218804364

is the Community Forest (HKm). One of the planning forms is the preparation of general work plans and annual work plans. Conclusions In the Aor Nakan Village, 91% of respondents chose to be involved in social activities as the highest form of participation, and 7% choose to contribute property as the lowest form of participation. In Kuta Tinggi Village, 91% choose to engage in social activities as the highest form of participation, and 11% choose to contribute property as the lowest form of participation. Whereas in Sibongkaras village, all or 100% of respondents choose to be involved in social activities as the highest form of participation. The level of participation in the three Forest Farmer Groups at the planning and implementation stages was high, reaching 100%. At the monitoring and evaluation stage in the Aor Nakan village 24% respondents with medium level of participation and 76% respondents with high level of participation, in Kuta Tinggi Village, 4% of respondents with low level of participation, 19% respondents with medium level of participation and 78% with a level of participation high participation. Whereas at the utilization stage of the results, in Aor Nakan Village 33% of respondents with medium level of participation and 67% of respondents with high level of participation, in Kuta Tinggi Village, 52% of respondents with medium level of participation and 48% of respondents with a high level of participation, in Sibongkaras Village, 32% respondents with medium level of participation and 68% of respondents with high level of participation.

218804364

Integrating Preventive Maintenance Planning and Production Scheduling under Reentrant Job Shop This paper focuses on a preventive maintenance plan and production scheduling problem under reentrant Job Shop in semiconductor production. Previous researches discussed production scheduling and preventive maintenance plan independently, especially on reentrant Job Shop. Due to reentrancy, reentrant Job Shop scheduling is more complex than the standard Job Shop which belongs to NP-hard problems. Reentrancy is a typical characteristic of semiconductor production. What is more, the equipment of semiconductor production is very expensive. Equipment failure will affect the normal production plan. It is necessary to maintain it regularly. So, we establish an integrated and optimal mathematical model. In this paper, we use the hybrid particle swarm optimization algorithm to solve the problem for it is highly nonlinear and discrete.The proposedmodel is evaluated through some simple simulation experiments and the results show that themodelworks better than the independent decision-makingmodel in terms of minimizing maximum completion time. Introduction In 1993, Kumar [1] proposed the third mode of production, reentrant production, which was different from common Job Shop or Flow Shop when he studied the semiconductor manufacturing process.Reentrancy means that the same job, at various processing stages, may wait for processing before the same device or the workpiece, at different stages of processing, visits the same device repeatedly.This feature makes the process even more intense when different jobs are processed in the same machine, which will undoubtedly make scheduling more difficult.With the rapid development of semiconductor industry, the reentrant Job Shop scheduling problem (RJSP) has gotten more and

58933189

more attention recently.What is more, semiconductor manufacturing equipment is extremely expensive.In the production process, an indeterminate fault might happen to these machines which will affect normal production planning.To improve the efficiency of the whole production system, equipment maintenance should also be considered at the same time.Combining production scheduling with preventive maintenance is worth discussing on RJSP.We propose an integrated and optimal mathematical model in this paper and use the hybrid particle swarm optimization algorithm to solve the problem.The proposed model is evaluated through simulation experiments and the results show that the model works better than the independent decision-making model in terms of minimizing maximum completion time. RJSP is more complex than JSP because the same job will visit the same machine more than once at different stages.The main solving method is the heuristic algorithm.An approach of artificial neural network is proposed to solve multidecision scheduling problems of semiconductor wafer fabrication [2].Topaloglu and Kilincli [3] proposed a modified shifting bottleneck heuristic for the reentrant Job Shop scheduling problem with makespan minimization objective.Zoghby et al. [4] built a model for reentrant Job Shop scheduling problem with sequence-dependent setup times and investigated the feasibility conditions for metaheuristic searches.Danping and Lee [5] gave a full picture of the research methodology for the reentrant scheduling problem.Two-level hierarchical production planning based on linear programming was proposed for semiconductor wafer fabrication in [6], which was proved to be better than an existing algorithm by simulation experiments.Gupta and Sivakumar [7] made a review on Job Shop scheduling techniques in semiconductor manufacturing.Mönch et al. [8]

58933189

discussed a complex Job Shop environment, which contains parallel batching machines, machines with sequencedependent setup times, and reentrant process flows, and a modified shifting bottleneck heuristic was used to solve this complicated problem. Although many achievements have been made in semiconductor manufacturing production scheduling, they are studied only from the perspective of the production department to optimize production scheduling.They ignore the fact that the semiconductor manufacturing workshop equipment maintenance department arranges a large number of preventive maintenance tasks so that they will affect the equipment production.What is more, the random failure of machines will interrupt production scheduling.It will generate a production scheduling conflict when we discuss production scheduling and equipment maintenance separately.Obviously, the related theoretical study of production scheduling does not reflect the actual status of the reentrant shop job scheduling very well at present. Equipment preventive maintenance is also important to improve production efficiency.Ma et al. [9] built an equipment maintenance scheduling model of semiconductor manufacturing and proposed a method which is called particle swarm optimization algorithm to solve the problem.Mosley et al. [10] compared different dispatching maintenance scheduling policies which can significantly affect system performance by using a discrete event simulation model of a wafer fabrication.Davenport et al. [11] presented a goal programming approach that incorporates both constraint programming and mixed-integer programming solution technologies.Yao et al. [12] proposed a two-level hierarchical modeling framework to solve preventive maintenance scheduling.At the higher level, there is a model for long-term planning, and, at the lower level, there is a model for shortterm PM scheduling.Based on the above

58933189

research, we can see that equipment maintenance can greatly improve production efficiency.However, connected equipment maintenance and production scheduling have more realistic significance. Although joint optimization is rarely studied on semiconductor production, many researches have been carried out in traditional production scheduling and equipment maintenance.Cassady and Kutanoglu [13] proposed an integrated model to solve the problem of production scheduling and equipment maintenance on a single machine.Da and Wang [14] built an integrated model about production scheduling and equipment maintenance under the Flow Shop and proposed a heuristic approach to solve this problem.Da et al. [15] have done some researches on batch scheduling and maintenance under the Flow Shop and established a joint optimization model.Kaihara et al. [16] proposed a new approach, which is based on Lagrangian decomposition coordination, to solve the problem of maintenance scheduling.As for the complex workshop environment such as JSP, studies on joint optimization are very few.It will consider multiobjective optimization measures when optimizing processing sequence and maintenance time at the same time.Some papers are always introducing other objective functions such as system availability in order to reflect the influence of maintenance decision-making on the reliability of the equipment [17][18][19]. On the whole, the existing research on joint optimization is preliminary; most of the researches focus on basic workshop types and mode of production.They do not consider the reentrancy of complex production workshop.Judging from the current research results, although joint optimization of production and maintenance has aroused more and more attention, studies on solving joint research of reentrant manufacturing production and equipment maintenance operations

58933189

are very rare.In this paper, we combine production scheduling with preventive maintenance planning and establish an integrated optimization model of scheduling and preventive maintenance with the goal of minimizing completion time. The paper is organized as follows.Section 2 is devoted to representing the integrated problem and building a mathematical model.Section 3 describes the process of hybrid particle swarm optimization algorithm.Section 4 presents the simulation results and comparison result.Section 5 discusses the conclusion and future research direction. Problem Formulation Standard JSP contains two subproblems: the machine selection problem and the operation sequencing problem.During the production, each job can only be processed on different machines only once.And each machine can process a job at the same time.The processing diagram is shown as Figure 1.The processing sequence on machines is 1→2→3→5.However, RJSP is more complicated and challenging than the standard JSP because it exhibits a reentrant situation, where different operations of the same job can be processed on the same machine.The processing diagram of RJSP is shown as Figure 2. The processing sequence on machines is 1→2→3→4→2→5.The job is processed on machine 2 twice.At the same time, if the machines fail during production, this will affect the whole production plan.So, we develop a joint mathematical optimization model to solve this problem in this paper.The objective of this problem is to minimize the maximal completion time of all the operations. 2.1.Assumption.RJSP and equipment preventive maintenance planning problem can be represented as follows: there are machines in a workshop and different jobs should be processed.It has been arranged which operation will

58933189

be processed on which machine before production.Different operations of the same job may be processed on the same machine not only once.Each machine may have different abilities.Some hypotheses considered in this paper are summarized as follows: (1) The processing time for each operation in a particular machine is defined. (2) A machine can only process one operation at a certain time (resources constraint). (3) Each job must be processed on one machine at a given time, so jumping the queue is not allowed. (4) All the jobs have the same probability to be scheduled at the beginning, which means different jobs have the same priority. (5) Each operation of each job is not allowed to be interrupted in processing time; that is, preventive maintenance operations must be before or after production. (6) All machines are new at the beginning of the production process. (7) Each machine recovers as new after the implementation of preventive maintenance. (8) The time of small repair and machine setup is ignored during the processing. Definition of Notations. The indices and decision vari- ables used in the model are as follows. 𝑂 is performed on machine in priority 0, otherwise 1, machine conducts preventive maintenance after finishing th procession 0, otherwise. (1) Maintenance Strategies. In the process of production, equipment maintenance is a basic work.Timely maintenance can restore equipment performance, fix the trouble, and effectively prolong the service life of equipment.So, it is one of the most important ways to ensure smooth production. For preventive maintenance, it can be roughly divided into

58933189

two categories.One is preventive maintenance which is based on machine reliability, and the other is periodic preventive maintenance. For preventive maintenance based on the reliability of the machine, when the reliability of the machine reaches the threshold set, it is time for equipment maintenance.Some scholars suppose the machines will return to the new state after maintenance in their studies.In fact, equipment performance is gradually degraded with decreasing machine age.Under the same maintenance conditions and maintenance time, the equipment degree of maintenance decreases with the increase of the maintenance times.So, it will take less and less time to get the reliability threshold.So, lots of scholars assume that the reliability threshold will reduce after maintenance.Periodic preventive maintenance is the most common method in industrial production.It means that when each machine achieves the maximum maintenance interval, it must be maintained in time.When there is small equipment failure, we assume that the maintenance time is negligible.So, we do not consider the failure rate of the machine.Thus, this article mainly uses the periodic preventive maintenance strategy. In practice, many companies perform PM at a fixed interval.This kind of fixed interval is according to the processing time when starting production.If we calculate time according to processing time, the machines will be maintained excessively because some machines may not process any job during the maintenance cycle.So, we adopt other methods to determine PM intervals.When we calculate time, we focus on the age of the machine.This method is used in many researches [17,18,20].For example, one job is processed on 2 machines and it has

58933189

4 operations.The processing time is as Table 1 shows.We assume preliminary maintenance cycle = 15. If we calculate time according to machine age, each machine just needs only onetime maintenance as depicted in Figure 3.However, each machine should be maintained twice if we calculate time according to processing time described in Figure 4.The increasing maintenance times will increase maintenance cost and delay production scheduling.So, the maintenance strategy based on machine age is more reasonable for our study. Mathematical Model 2.4.1.Objective.The main goal in this paper is to minimize the makespan ( max ), which is the cumulative time to complete all operations on all machines with maintenance. (2) Jobs' Starting Production Time on Machines.The starting production time of the th job, which is produced on the th machine times, is dependent on many factors.They are as follows: (1) completion time of previous jobs on jth machine: (2) completion time of the previous operation of the ith job on another machine (3) time for implementing PM.In our model, PM is regarded as perfect, which means the machine will be renewed to an as-good-as-new status.We assume that the initial age of the machines is zero.Then, according to the machines' age and fixed maintenance cycle, we can decide whether PM should be conducted.Hence, the completion time of the current job can be calculated as follows: ( The completion production time of the ith job, which is produced on the jth machine times, is calculated as follows: Due to reentrancy, the same kind of job in different stages will wait

58933189

to be produced on one machine.So, we should decide which one will be produced firstly.The calculation is as follows: ∀, ℎ, ∀ , ℎ , ∀, = 1, 2, . . ., . (7) Hybrid Particle Swarm Optimization Algorithm 3.1.Workflow of the Proposed HPSOA.With high nonlinearity and discreteness, the reentrant Job Shop scheduling problem is a strongly NP-hard problem.It is very hard to be solved by traditional optimization methods within an acceptable time.In recent years, many scholars have been committed to using intelligent optimization algorithms to solve this problem, such as genetic algorithm, ant colony algorithm, and particle swarm optimization algorithm.General particle swarm optimization algorithm is to find the optimum value by tracking individual best value and group best value [21].Although this algorithm is simple and its convergence rate is very fast, it will be trapped in local optimum and cannot jump out with increasing iteration times.Hybrid particle swarm algorithm discards the method of tracking the extreme to update the particle's location in the traditional particle swarm optimization algorithm.It introduces crossover and mutation in the genetic algorithm.The way of searching the optimal value is to find the best particle after crossing operation and mutation operation.So, the hybrid particle swarm optimization algorithm (HPSOA) [21] is mainly applied to solve the integrated problem in this paper. Workflow of the proposed HPSOA is shown in Figure 5.All the details of the proposed HPSOA will be described in the following subsections.The overall procedure of the proposed approach is described as follows. Step 1. Initialize the number of particles, the maximum

58933189

number of iterations, learning factors, and so on. Step 2. Initialize the particle's initial position and velocity and calculate the fitness value. Step 3. Initialize the particle's personal best and global best. Step 4. Calculate the value of inertia weight and update the particle's velocity and position. Step 5. Calculate the fitness value of particles and then sort order according to the fitness value. Step 6. Improve particles by crossover and mutation operation. Step 7. Recalculate the fitness value. Step 8. Obtain updated personal best and global best.Step 9.If the termination condition is met, output the optimal solution.Otherwise, execute the loop body Steps 4 to 8. Encoding. In our encoding, the particle is composed of two-dimensional vectors.The job sequence in the first dimension decides the job's processing sequence. The second dimension is a set of random numbers which can decide particles' velocity and position.The range of those random numbers is [0, 1].It has been arranged which operation of each job will be processed on which machine before production.If job has operations, it will generate .For example, if there are 2 jobs, each will have 3 operations.The job sequence is And generate a set of random numbers ranging from 0 to 1.The primary particle is [1 1 1 2 2 2; 0.8147 0.9058 0.1270 0.9134 0.6324 0.0975].This set of random numbers is sorted in an ascending order.And the job sequence will change accordingly.The real particle is [2 1 2 1 1 2; 0.0975 0.1270 0.6324 0.8147 0.9058 0.9134].This means that the first operation of job 2 is

58933189

processed firstly and then the first operation of job 1 is processed and so on. Crossover. If the particle has reached the condition of crossover, we choose the particles to cross with the best personal particle and best group particle.The method of crossover is integer crossover.At first, we choose the position of crossover randomly.And then the selected one crosses with the personal best or group best.For example, if we choose the second position and fourth position to cross, the operation will be depicted as follows. Individual After crossing, the amount of processing of some jobs will increase or decrease.In this situation, we will take necessary actions to make sure the amount of processing meets the conditions.In the crossover process, we will choose the best individual to update the particle swarm according to the fitness value. Mutation. In the mutation operation, two positions are changed with each other in the same individual.At first, we choose the mutation position, position 1 and position 2, and then change the mutual position.We just change the position of the job not the particle's position.For example, if we change the second position and fourth position, it will be described as follows: Simulation and Discussion 4.1.Simulation.In real semiconductor manufacturing scheduling, the production process is very complex.Each job needs to go through hundreds of operations.In order to validate the superiority of joint optimization model and the effectiveness of hybrid particle swarm algorithm, we compare it with the independent decision-making model. In the example, there are 4 jobs that are processed on 5 machines in

58933189

the system.Some machines will produce the same job more than once.In order to reflect the competition of similar products at different stages on the same machine, we assume that job 1 and job 2 are the same products.Job operation information is depicted in Table 2.In the array, the first number presents processing machine and the second number presents processing time. Maintenance has degraded effect in the whole production process.Kubzin and Strusevich proposed a linear degradation function to calculate PM time [22].In this example, we also use = + Δ to calculate PM time.We suppose = 0.5.In this paper, the hybrid particle swarm optimization algorithm is used to minimize makespan and MATLAB 2010b is used to realize the algorithm.The algorithm parameters are set as follows: the particles' initial number is 200 and the number of loop iterations is 500. is the crossover probability in . max equals 1 which is the maximum crossover probability. min equals 0.5 which is the minimum crossover probability. max which represents the number of generations equals 200. is the mutation probability in . max equals 0.1 which is the maximum mutation probability. min equals 0.01 which is the minimum mutation probability. 4.1.1.Integrated Optimization Strategy.By using hybrid particle swarm optimization algorithm, the optimal solution is 392.5.The Gantt chart of integrated optimization model is shown in Figure 6.In the figure, the horizontal axis is the time axis, the vertical axis is the machine order, the same color box in the figure represents the same job, and the red box denotes equipment maintenance. Independent Decision-Making

58933189

Strategy. In actual production, the production department is responsible for the production scheduling, and the equipment maintenance department just makes equipment maintenance plan.So, their work is separated.The maintenance department calculates the best maintenance cycle, and the production department calculates the best production scheduling, and then the production department arranges a production plan. In the study, we assume that the production sequence is fixed.And if the machine's cumulative processing time is more than maintenance cycle, the machine needs to be maintained right now.Used by the same algorithm, the optimal solution is 411.5.Its scheduling Gantt chart is shown in Figure 7. Discussion of Result. Compared with independent decision-making, the result is improved by 4.62%.From the Gantt chart, we also can get that the maintenance time of joint decision-making is less than of independent decisionmaking.So, this will decrease the maintenance cost in some ways.In order to prove the superiority of the combined optimization model, we simulate a few more complex numerical experiments, and the simulation data can be checked in Supplementary Materials available online at https://doi.org/10.1155/2017/6758147.The final results are shown in Table 3.On the whole, the joint decision-making model performs better than the independent decision-making model in solving the problem of production scheduling and maintenance plans. Evaluation of Hybrid Particle Swarm Optimization Algorithm (HPSOA). Genetic algorithm (GA) is a stochastic search and optimization method which is based on natural selection and genetic mechanism [23].In recent years, the genetic algorithm had great potential in solving complex optimization problems and it has successful application in the industrial engineering field.In the area

58933189

of production scheduling, this algorithm is also very important.There are many scholars using this algorithm [24][25][26][27]. To verify the validity of HPSOA, we make some tests on different examples compared with GA.In the testing, we make comparison with GA which can also be used to solve this problem.And the results indicated that HPSOA is better than GA to solve this problem. At first, we analyze the example of 4 jobs and 5 machines we talked about above.The final result is 403.5 which is worse than HPSOA.Its scheduling Gantt chart is shown in Figure 8. The machines' reentrant times will affect dispatching, so reasonable arrangements for these machines play a big role in improving production efficiency.So, we give two more complex examples to verify the superiority of our method by increasing reentrancy times.One is 5 jobs/7 machines with 3 times reentrancy.The other is 5 jobs/7 machines with 5 times reentrancy.They are all solved using GA and HPSOA.The results are shown in Table 4.Although the running speed of HPSOA is slower than GA, the results are better especially with the number of operations and reentrancy times increasing.So, HPSOA is preferable to be used to solve our problems. Conclusions and Future Research In actual production, production scheduling and preventive maintenance are closely linked and interactive.In this paper, we build an optimal model of reentrant manufacturing scheduling and equipment maintenance.We adopt fixed maintenance cycle with penalty function strategy and solve the problem effectively by the hybrid particle swarm optimization algorithm.By comparison with GA, this method is preferable to solve the

58933189

combined optimization problem.We make lots of assumptions to simplify the complex problem for our study.However, they will not exist in practical production.Firstly, equipment failure is difficult to avoid, so the machines' minor repair time should not be ignored.What is more, equipment age is changeable after maintenance in reality.At last, we focus on only one single objective in the study.However, cost is an important factor to consider which includes production cost, preventive maintenance cost, minimal repair cost for unexpected failures, and tardiness cost.As a consequence, future efforts are still needed to solve this composite problem in reentrant Job Shop production scheduling. Figure 6 : Figure 6: The Gantt chart of integrated optimization model. Figure 7 : Figure 7: The Gantt chart of independent decision-making. Figure 8 : Figure 8: Job processing and machine maintenance by GA. 0 : the initial age of machine k. : preventive maintenance time on machine (if the maintenance interval iswithin the fixed time , then = ; if not, = + Δ, where Δt is the over time and is a parameter). : processing time of th operation processed on machine . : starting time of th operation processed on machine . : the age of machine starting to process th job. : the operation of job processed times on machine . : processing time of . : completion time of operation . : starting time of . Table 1 : Job processing time on machines for the example. Table 2 : Job processing time on each machine (time units). Table 3

58933189

: Comparison of independent decision making and integrated optimization. Table 4 : Comparison of GA and HPSOA for complex problems.

58933189

LncRNA-Top: Controlled deep learning approaches for lncRNA gene regulatory relationship annotations across different platforms Summary By soaking microRNAs (miRNAs), long non-coding RNAs (lncRNAs) have the potential to regulate gene expression. Few methods have been created based on this mechanism to anticipate the lncRNA-gene relationship prediction. Hence, we present lncRNA-Top to forecast potential lncRNA-gene regulation relationships. Specifically, we constructed controlled deep-learning methods using 12417 lncRNAs and 16127 genes. We have provided retrospective and innovative views among negative sampling, random seeds, cross-validation, metrics, and independent datasets. The AUC, AUPR, and our defined precision@k were leveraged to evaluate performance. In-depth case studies demonstrate that 47 out of 100 projected top unknown pairings were recorded in publications, supporting the predictive power. Our additional software can annotate the scores with target candidates. The lncRNA-Top will be a helpful tool to uncover prospective lncRNA targets and better comprehend the regulatory processes of lncRNAs. INTRODUCTION The long non-coding RNAs (lncRNAs) are not translated into proteins and can span over 200 nucleotides. 1,2More than 16,000 lncRNAs were reported in the Homo sapiens, as revealed in the human GENCODE project. 3Although lncRNAs are less characterized than protein-coding genes, their annotations may impact downstream research. 4By controlling target gene expression levels, lncRNAs have various activities, influencing motility, invasion, differentiation, apoptosis, and proliferation. 5Researchers have linked lncRNA dysregulation and mutations with a greater propensity for complex disorders such as cancer. 6,7inding new lncRNA relationships with targets as genes may be crucial for developing tailored therapies or discovering disease mechanisms.However, as lncRNAs may affect genes 8,9 in various

264084904

ways, 10 finding new relationships can be difficult.In vivo validation tests are timeconsuming and labor-intensive; thus, the in silico screen has great potential.In silico experiment relies on prior knowledge such as databases.Fortunately, researchers have curated databases concerning lncRNA, such as LNCipedia 11 for lncRNA sequences, lncRNAfunc 12 for lncRNA functions, LncRNADisease 13 for lncRNA illnesses, and LncRNA2Target 14 and LncTarD 15 for lncRNA-target pairings.Those datasets provide prerequisites for in silico experiments. Previously, researchers have conducted in silico experiments on lncRNA.For instance, to predict latent lncRNA-disease interactions, Xuan et al. 16 combined heterogeneous networks with graph convolutional networks (GCN) and convolutional neural networks (CNN).DeepMNE 17 merged the disease similarities derived from disease semantic information, functions, Gene Ontology terms, and lncRNA similarities originating from sequences and expression files, using known lncRNA-disease relations to predict new lncRNA-disease associations. Other methods directly predict the relationships between lncRNA and other molecules.For instance, Huang 18 designed a GCN autoencoder to predict lncRNA-microRNA (miRNA) relations from molecular networks.Fukunaga 19 collected tissue-specific expression profiles and subcellular localization information to predict potential lncRNA-mRNA relationships.A recent study that employs deep learning to reveal potential lncRNA-gene pairs is DeepLGP, 20 which collects expression and genomic location features and constructs positive pairs based on LncRNA2Target 14 and negatives from random lncRNA-gene pairs.Additionally, GCN is applied to extract features, while CNN is employed to make predictions farther down the line.IRDL 21 integrated sequencing, gene expression, and chromatin accessibility to identify divergent lncRNAs with target genes.The LPI-deepGBDT 22 takes gradient-boosting decision trees and features from Pyfeat and BioProt for potential

264084904

lncRNA-protein relationship prediction.DNNMC 23 proposes a lncRNA-protein-coding gene (PCG) computational method combining deep ll OPEN ACCESS neural networks and inductive matrix completion using multi-omics data and known association.GAE-LGA 24 leverages multi-omics features of lncRNA and PCG as the source for the similarity calculation and applies the graph-autoencoder for potential relation interference.Most of those methods considered expressional profiles, multi-omics data, or location of gene and lncRNA as features. Nonetheless, few considered the functional mechanism, such as lncRNA playing a role as ce-RNA to regulate the gene. 25,26For example, lncRNA-PNUTS, an alternatively spliced lncRNA from PNUTS premature mRNA, contains seven binding sites to miR-205, enhancing ZEB1 and ZEB2 mRNAs by reducing miR-205-binding ability. 27The miRNA can regulate gene post-transcription by binding to the 3 0 UTR of genes. 28ncRNA can also provide binding sites for miRNA.Thus, building predictive models directly from lncRNA and gene 3 0 UTR sequences seems practical.The next thing is to select predictive models. In other bioinformatics fields, deep learning has had considerable success, such as protein structure prediction, 29 inference of miRNA-gene links, 30 compound-protein interactions, 31 cancer survival analysis, 32 and cell fate. 33However, deep learning is likely to show considerable variance, especially when using tiny datasets with high dimensions, such as omics characteristics from biological features, 34 which deteriorates the model's generalization ability.The model's capacity for generalization demonstrates its flexibility in responding to new data.However, many previous deep learning models, such as our previous work, only considered the area under the curves (AUC) values inside one dataset.Furthermore, when choosing the negative samples,

264084904

how the random seeds may influence the results is merely explored.That lack of generalization verification demands independent datasets and more experiments on different random seeds. Ensemble approaches are an excellent solution to increase model generalization.Combining the output from many deep-learning models can improve the model's generalizability. 35Mixed deep-learning and machine-learning models may further improve generalization capabilities. 36Ensemble techniques have been employed in many link prediction tasks of bioinformatics applications, including Cas9 off-target, miRNA-target, microbiome-drug, and microbiome-disease.For instance, Zhang et al. 37 used ensemble AdaBoost to predict CRISPR-Cas9 off-target activities by combining characteristics from synergizing methods, conservations, and chromatin annotations.EnANNDeep 38 is an ensemble method aggregating AkNN, DNN, and gcForst results for scoring lncRNA-protein pairs.The features of which are extracted from protein sequences and structures.To predict miRNA-target correlations, SRG-vote 39 integrated several long short-term memory models trained with features from different aspects.Long et al. 40 proposed an ensemble graph network with three models and several input sources to identify potential relationships between medications and microorganisms.Chen et al. 41 made soft voting of four CNN models for human disease-microbiome prediction.These studies indicate that integrating models may have better performance.However, those methods only considered ensembles inside one dataset with different models, not considering ensembles with different negative sampling random seeds and validating them on the independent datasets. Inspired by this, we designed lncRNA-Top, the controlled deep-learning approach (random forest (RF)-CNN ensemble) to predict lncRNAgene relationships leveraging sequence-based characteristics inspired by the lncRNA regulatory mechanism.In the meantime, we have retrospected and revisited some spots and details that are easily ignored

264084904

in deep-learning tasks, including negative sampling with varied random seeds, different types of cross-validation, results confidence by the statistical calculation, varied independent datasets, ensemble verification on independent datasets, and differentiated metrics.Except for AUC, we also introduced other metrics, such as area under precision-recall (AUPR) curve and our defined [email protected] one model can get a higher AUPR, it can control the false positive rate (FPR), 42 which also means that the top-predicted pairs are more likely to be the ''real-positive'' pair.These metrics help us polish our ensemble policy when predicting the actual scenario. The contributions of lncRNA-Top are listed as follows. Verification using three different approaches on each dataset Three types of verification results for each sub-model, such as RF and CNN, on different datasets are shown in Figure 1.Rank-sum tests were leveraged to calculate the adjusted p value with Bonferroni correction.Except for the type-1 and type-3 AUPR in the low-throughput constructed (LC) dataset, all datasets show that CNN performs better than RF with adjusted p values less than 0.05.From the results, some interesting conclusions can be drawn, such as: Integrating the IT and LC datasets may improve the model's performance and lower error bars; leveraging more training data (90% of data instead of 80%) results in higher performance in type-3 than in type-1 and type-2. Figures 1D and 1H demonstrate the overall performance of merging the results of type-3 verification of AUC and AUPR.We can see from the images that for the 10-fold cross-validation, RF's results are less variable than CNN's, demonstrating that deep learning exhibits

264084904

significant variance, which means that the deep-learning models (CNN) are more easily affected by the random seeds.The results further indicate that, when conducting deep learning experiments inside one dataset, it is necessary to repeat the experiments from the negative sampling. Comparison with the SOTA methods Although our method is specially designed for the lncRNA-gene regulation relations, given that we share the same training datasets with the SOTA method DeepLGP, 20 we could still compare our AUC/AUPR with theirs.We also leveraged the 10-fold cross-validation.Details of the results are shown in Figures 2A-2D. From the results, we could see that the RF with Poly-kernel-PCA-transformed features could get significantly higher results (adjusted p value with Bonferroni correction % 0.05) in both the LC and high-throughput constructed (HC) datasets.DeepLGP leverages GCN and CNN to discover potential relationships between lncRNA and genes, a deep-learning-based model.We only leveraged our non-deeplearning algorithm and got higher AUC/AUPR, demonstrating that our mechanism involved sequence-based methods that are working. One rationale is that we utilized more advanced features than the SOTA approaches; and another is that we created more gene and lncRNA features.We further explored the feature generated from different lncRNA-related predictive models.A detailed method comparison can be found in Table 1.We compared those features in two steps.The first step is leveraging LPI-generated features in our predictive framework; the results are shown in Figure 2E.The second step is to compare the performance of different features under the framework GAE-LGA.Those features include original multi-omics features, LPI's features, and our features.The results of different features under GAE-LGA can

264084904

be found in Table 2. From the figure, we could find that the features generated by the lncRNA-Top can achieve better performance for all sub-models and metrics.Table 2 demonstrates that the original multi-omics features are better than LPI's generated features but inferior to ours We first experimented in the adjusted network (after removing the lncRNA/gene we don't have).We acquired a benchmark value of GAE-LGA, named GAE-LGA (Adjusted), as the adjusted network is smaller than the original one.The metrics calculated are slightly inferior to the original paper's description.Then, we replaced the original multi-omics features with LPI features and our features and reconducted experiments again.The bold value corresponds to the best performance method for each metric.The results show that our features can outperform most of the metrics.LPI is designed to predict lncRNA and protein relationships and provides feature generators from sequences.GAE-LGA is a graph-based method proposed for lncRNA and PCG (protein-coding gene) relation prediction.In the framework of GAE-LGA, features were leveraged for similarity calculation.We compared our method in their framework by replacing its original multi-omics features with LPI generated and our features. Transfer verification for independent datasets The results (sorted by the models) of the transfer verification among independent datasets are shown in Figure 3.We compared the mean value of AUC/AUPR of RF, CNN, RF-CNN, and our controlled ensemble methods, RF-CNN ensemble.As can be seen from the picture of Figures 3A and 3B, the RF-CNN-ensemble (last four rows) can get lighter color in both AUC (yellow) and AUPR (blue), indicating that aggregating the scores predicted from different random-seeds

264084904

generated training sets can increase the metric of AUC/AUPR. Ensemble composition As ensemble policy from different training random seeds can increase the overall performance, the next thing is to figure out the optimal composition of the ensemble.The LCIT-HC is more likely to be the actual scenario, where the pairs to be predicted are enormous, and the training dataset is limited.Thus, we further generated ROC and PR curves of the CNN and RF ensemble on the LCIT training set.We also introduced the sub-model-only results as a comparison.The results are shown in Figure 4. Prediction ability comparison To further assess the RF ensemble, CNN ensemble, and RF-CNN ensemble's capacity for prediction, we trained models on the LCIT dataset.All pairs of lncRNA and genes were taken into consideration.The precision@K is leveraged for the comparison.The results are listed in Table 3. Here, we select k between 10 and 100.The results demonstrate that the RF-CNN ensemble could get much higher precision@K in all scenarios, which denotes better predictive performance. Parametrical analysis In the framework of lncRNA-Top, we have plenty of parameters that might contribute to the model robustness.To further explore those factors, we conducted a parametrical analysis.The variables include kernels and dimensions of Kernel PCA when transforming the features, different sub-machine-learning models (Figure S1), negative sampling rate (Figures S2-S4), and random seeds (Figure S5; Table S1) for cross-validation and ensemble. RF-CNN ensemble zoomed in comparison As shown in Figure 5, the RF-ensemble model seems to be the best option as it can achieve higher true positive rate or grow fast when

264084904

the FPR rises in all predicting HC tasks.However, when we further screen the predictive results from the RF ensemble with all pairs of 12417 lncRNA and 16127 genes, their top results frequently tend to be indistinguishable.(RF-ensemble predicted pairs shared identical scores, resulting in the same rank.)Thus, by adding the CNN scores, the controlled deep-learning algorithms (RF-CNN ensemble) may produce more separable top outcomes with slight deterioration to AUC/AUPR but a significant increase to the top-predicted results (precision@K).Another possible reason why the RF-CNN ensemble is better in real predictive situations is that the RF-CNN ensemble is the aggregate results of six models.In comparison, RF ensembles and CNN ensembles only contain three models. Top predictive results To further illustrate the predictive ability of lncRNA-Top, we searched the literature for the highly predicted lncRNA-gene relationships.The top ten predicted results (those that have appeared in the training set are removed) are shown in Table 4. Figure 4. Ensemble composition The figure described the ROC and PR curves when using LCIT to predict the HC dataset.This project also leverages the small dataset to predict the large dataset, which is the closest to the actual scenario.In the first picture, RF-only is barely overlapped with RF-CNN-only, and RF-ensemble is barely overlapped with the RF-CNN ensemble.The ROC and PR curves are drawn based on the concatenated label and scores of the independent dataset (for example, if we leverage the LCIT_rs1 to train the model and generate a score for the HC_rs2, rs3 sample, we would concatenate scores and labels to generate RF-only or CNN-only

264084904

ROC curves).The independent sample score is the sum of different random seeds trained models for ensemble methods.As can be seen from the picture, the performance of ensemble methods like RF-ensemble and CNN-ensemble are increased compared with their sub-model.Also, the RF ensemble can get the highest AUC and AUPR in this task.The RF-CNN-ensemble ranked second in this task.The regulatory mechanism that inspired us to construct this predictive model is that the lncRNA can regulate miRNA (microRNA) while miRNA can regulate genes.Thus, lncRNA can be a competitor to some genes that share the same target miRNAs.Our previous publications 30,39 demonstrated that modelling miRNA-gene relationships between miRNA and 3 0 UTR gene sequences worked.Inspired by this, in this model, we also constructed our predictive models using 3 0 UTR gene sequences. Interestingly, although no miRNA information (as we directly modeled on lncRNA sequence and gene sequence) was introduced to our method, they were mentioned in most publications.We list the miRNA in the table and find that five out of the top 10 lncRNA-gene relations predicted by our framework are related to miRNA.Here are some examples: the miRNA miR-370-3p can target gene mitogen-activated protein kinase (MAPK1), and therefore, lncRNA taurine-upregulated gene 1 (lncRNA TUG1) could reduce the level of functional miR-370-3p and facilitate MAPK1 expression. 43The elevated level of miR-197 in cells treated with lipopolysaccharide (LPS) was inhibited by transfection with TUG1, which can sponge miR-197 to enhance the level of p-MAPK/MAPK, thereby inducing autophagy, indicating the upregulating of TUG1 might inhibit the increase of miR-197 and enhance MAPK. 44SNHG1

264084904

and MAPK1 were significantly upregulated, while miR-125b-5p was downregulated in the MPTP-induced PD mouse and MPP+-induced PD cell models. 45SNHG1 competitively binds to the miR-221/222 cluster and indirectly regulates the expression of cyclin-dependent kinase inhibitor 1B (CDKN1B/p27). 46NEAT1 forms double-stranded RNA with miR-222-3p, thus limiting miR-222-3p's binding with CDKN1B, which indicates that the upregulate of NEAT1 will reduce the miR-222-3p.Thus, CDKN1B would be upregulated. 47lncRNA NEAT1 was upregulated while miR-27a-3p was downregulated in SH-SY5Y cells, and CASP-3 protein and its lytic cell protein were upregulated. 48Neat1_2 (a transcript variant of NEAT1) competitively binds to miR-129-5p and prevents miR-129-5p from decreasing FADD, CASP-8, and CASP-3 levels, ultimately facilitating TEC apoptosis.Neat1 binding to miR-129-5p, then the level of CASP-3 is increased. 49We further marked the direction of regulation that the papers indicated, and these cases Here, the FPR is cut off by 0.01.As can be seen from the picture, when the false positive rate (FPR) is small and begins to grow, the true positive rate (TPR) rises quickly.Considering we have 12417 lncRNAs and 16127 genes, the total pair is 200,000,619.Thus, even if a small part of the TPR increases, it can significantly increase the predictive ability.The ⬆ means the molecule is upregulated, ⬇ means the molecule is downregulated according to the publications.indicate that the lncRNA and gene are regulated in the same direction.If the lncRNA is upregulated and the corresponding gene is upregulated, the target miRNA is sponged and downregulated.The top-predicted lncRNA-gene pairs further indicate that building a predictive model based on the lncRNA sequences and gene 3

264084904

0 UTR sequences directly can reveal authentic gene transcriptional regulations by lncRNA through mechanisms such as competitive endogenous RNA.We also marked the year of publication.Those recent publications are not included in any dataset demonstrating lncRNA-Top's predictive performance. We tested the prediction of the case study dataset.Results can be found in Table S2. In this study, we proposed lncRNA-Top, controlled deep-learning approaches that predict potential lncRNA-gene regulatory relations inspired by regulatory mechanisms.The overall workflow of lncRNA-Top is illustrated in Figure 6.The usage and purpose of different datasets in this manuscript are shown in Table 5.Details of the constructed datasets leveraged for machine learning model training can be found in Table 6.We explored the influence of random seeds, ratios of negative sampling, different cross-validation, and transfer verification among datasets and conducted a multi-dimensional analysis of the predicted results.The controlled deep-learning approaches hybrid ensemble datasets and results of machine-learning and deep-learning models, increasing the AUC/AUPR/Precision@k of the predictive method.The case study denotes that our suggested approaches can accurately identify those lncRNA and gene regulatory relationships with substantial evidence.The code, features, software, and website are provided. Limitations of the study In our framework, we only considered the sequence-based feature extractors.However, there are many features from other domains, such as multi-omics and graph-based features.Introducing those features might contribute positively to the performance.We will explore more features in the next version of lncRNA-Top. STAR+METHODS Detailed methods are provided in the online version of this paper and include the following: Case study datasets 10 Extracted from, 10 the Independent dataset for 14,15 the

264084904

lncRNA sequences and 3 0 UTR complementary sequences from genes as input data for feature extraction.The LncRNA2Target v2.0 14 contains low-throughput and high-throughput datasets.High-throughput datasets contain more genes but fewer lncRNAs than lowthroughput datasets.We constructed three independent datasets from the low-throughput data (LC), the high-throughput data (HC), the lncTarD 15 dataset (IT), and the union set of LC and IT (named LCIT).Another independent dataset is manually curated from the literature 10 and named a Case study dataset.We filtered the datasets in advance to include unique lncRNA-gene pairs to avoid AUC/AUPR inflation.The case study dataset overlaps HC, LC, IT, and LCIT differently, as depicted in Figure 6G.Details of the constructed datasets can be found in Table 6. Negative sampling with random seeds We preserved the positive pairs and randomly chose the negative pairs to build embedding datasets.The random seed is noted after the dataset.For example, If we use the random seed 'one 'to generate the negative samples for the LC dataset, we remark them as LC_1.For our final ensemble predictor, we trained RF and CNN on the LCIT_1,2,3, tested it on the HC_1,2,3, and verified the top 100 predicted pairs with precision@k. Feature extraction We leveraged the iLearn platform 50 and the k-mer for sequence-based features.The iLearn platform is an open-sourced sequence feature calculation platform (software).We can generate a series of sequence-based features by inputting DNA, RNA, or protein sequences into the platform.We first generated all the features that do not need alignment in advance with the default parameters.Those features include Pseudo k-tuple composition (PseKNC), Dinucleotide-based Auto

264084904

Covariance (DAC), and Series correlation pseudo dinucleotide composition (SCPseDNC) et al.The k-mer is one of the rudimental features for sequence analysis.It contains essential information such as the statistical distribution, palindromic clips, and possible motifs.Taking "3-mer" as an example, assuming we have a sequence of "GTAC," then applying a sliding window to the sequence from right to left, such as "GTA" and "TAC."The count of those "k-mer" is the feature of one sequence.Herein, we set k from three to eight to guarantee variety but not the sparsity of the features.Features details can be found in Table S3. Kernel principal component analysis After the feature generation, the total feature dimensionality is exceptionally high (more than 84,000).If we build our machine-learning model directly and integrate all these characteristics, it is not computationally efficient.Thus, we applied feature transformation algorithms.The original feature space is usually non-linear separable.Thus, we applied Kernel Principal Component Analysis (Kernel PCA) with the poly kernel to extract the most variance-explaining features from the initial feature space.Mathematically, we denote the original features as x i .Applying a high-dimensional non-linear feature transformation ø to the data x i .We get: If we assume v is a linear combination of a high-dimension vector: Where a i are learned weights from the component of v. Thus, for a new point x à , the coefficient is based on the data points' similarity.The kernel PCA coefficient w for v is calculated as: Here, k is the p-order polynomial kernel function: kðx; x i Þ = ½ðx$x i Þ+1 p (4) In this

264084904

study, we selected 4096 features for the final prediction due to their overall good performance among different datasets. Constructed embedding datasets and metrics The lncRNA-gene pairs from the original datasets 14,15 were positive samples.The negative samples were randomly selected with random seeds. 20The number of negative samples is equal to the number of positive samples.We generated the constructed datasets several times to increase the robustness and calculate metrics statistics.The merge of positive and negative samples was regarded as the constructed dataset.We concatenated the lncRNA and gene sequence-based transformed features for each positive/negative lncRNA-gene pair to build embedding datasets and append them with label 1/0.Assuming F i;j as one row of lncRNA i and gene j, and L i;n and G j;n represent the n-th value of embedding, then the F i;j can be represented as: F i;j =  L i;0 ，L i;1 :::L i;n ; G j;0 ; G j;1 :::G j;n à ; (Equation 5) each pair of lncRNA and gene ensemble would have individual scores derived from several models.They would all be added together to determine the final score.The score and rank for each lncRNA/gene can be retrieved by our software published on our web server. QUANTIFICATION AND STATISTICAL ANALYSIS In this manuscript, Mann-Whitney-Wilcoxon tests (M.W.W. in figures, also known as rank-sum tests) were conducted to compare the performance of each feature extraction or prediction model.For cross-validation, each fold's results (AUC/AUPR values) would be regarded as minimal operation elements for statistical tests.The mean value is also calculated to show the overall performance of

264084904

each model. Figure 1 . Figure 1.Verification using three different approaches on each dataset (A-C) The barplot of AUC values inside IT, LC, and LCIT datasets with adjusted p value was annotated using the rank-sum test with Bonferroni correction using type-1, type-2, and type-3 methods.(D) The violin-swarm plot of AUC for type-3 10-fold cross-validation.(E-G) The barplot of AUPR values inside IT, LC, and LCIT datasets.(H) The violin-swarm plot of AUPR for type-3 10-fold cross-validation, each white point indicates one experiment with different sampling random seeds and 10-fold cut random seeds.Mann-Whitney-Wilcoxon tests were conducted, and p-values were calculated to compare the performance of each feature extraction or prediction model.The figure's error bar indicates a specific metric's distribution range. Figure 2 . Figure 2. Comparison with the SOTA method (A and B) The violin-swarm plot of AUC and AUPR of our methods, when compared with the SOTA method DeepLPG, in the low-throughput constructed (LC dataset) dataset, 1, 2, 3 here mean the negative sampling random seeds.Adjusted p value with Bonferroni correction is annotated.(C and D) The violin-swarm plot of AUC and AUPR of our methods, when compared with the SOTA method DeepLPG, in the high-throughput constructed dataset (HC dataset).(E) The barplot indicated the performance of features from LPI and ours.The x axis is the models and metrics.We introduced AUC, AUPR, F1, and MCC in this experiment.Mann-Whitney-Wilcoxon tests were conducted, and p-values were calculated to compare the performance of each feature extraction or prediction model.The figure's error bar indicates a specific metric's distribution range. Figure 3 . Figure 3.

264084904

Transfer verification for independent datasets (Heatmap) (A and B) Heatmap of average AUC/AUPR of different methods and training-testing sets.Each cube zipped one machine-learning method, one training dataset, and an independent testing dataset in the heatmap.The AUC/AUPR values are marked with light to dark colors, denoting the value from large to small.Every four rows indicate models or ensemble methods, each column indicates each independent dataset for testing, and each row indicates model-trained datasets, including HC, LC, IT, and LCIT. Figure 5 . Figure 5. RF-CNN ensemble zoomed in comparison (A and B) Zoomed in ROC curves of (A) LC predicts HC, (B) IT predicts HC, and (C) LCIT predicts HC.Here, the FPR is cut off by 0.01.As can be seen from the picture, when the false positive rate (FPR) is small and begins to grow, the true positive rate (TPR) rises quickly.Considering we have 12417 lncRNAs and 16127 genes, the total pair is 200,000,619.Thus, even if a small part of the TPR increases, it can significantly increase the predictive ability. Figure 6 . Figure 6.The overall workflow of lncRNA-Top The workflow can be divided into four parts.The yellow parts denote the feature generation, and the green part reveals the dataset for downstream works.The reddish part indicates the ensemble model we take for the prediction, and the bluish part illustrates the verification we have done to validate our models.(A and B) lncRNA/gene 3 0 UTR sequences examples.(C) Generation of iLearn features from the iLearn platform.(D) Generation of k-mer (k = 3 to k = 8) features.(E) LncRNA and

264084904

gene feature transformation by the KPCA with poly kernels.(G and H) Using the LC, IT, HC, and Case_study datasets to construct four embedding datasets using the previously generated features (with different random seeds).(I) RF model trained by datasets.(J) CNN models trained by datasets.(K) Validation methods (three types) inside each dataset and the transfer verification among each dataset.(L) Controlled deep-learning strategy, using different random seeds to construct a dataset and train sub-models.We predict the final results with RF and CNN models (controlled deep learning), differentiating the final scores and improving predictive performance.(M) Software and web server demo. Table 2 . Results of GAE-LGA with different features as input Table 1 . Comparison with the SOTA method Table 3 . The prediction ability of the different ensemble strategies Table 4 . Top unknown lncRNA-target relations predicted by lncRNA-Top Table 6 . The details of the constructed datasets Table 5 . Datasets and their usage in our method

264084904

Prevalence of compassion fatigue, burnout, compassion satisfaction, and associated factors among nurses working in cancer treatment centers in Ethiopia, 2020 Background Overuse of compassion for those under the care may threaten their professional life. In Ethiopia, there is limited study on the consequences of compassionate care. Therefore, the study assessed the effects of compassionate care among nurses. Objective To quantify the prevalence of compassion satisfaction, burnout, compassion fatigue, and associated factors among Nurses. Methods Institution-based quantitative cross-sectional design was conducted in five randomly selected public hospitals in Ethiopia, from May to April 2020. All the nurses who were working in the cancer treatment centers of the five hospitals were included in the study. Data were collected using a standard self-administer structured question using the Professional Quality of Life Scale (PROQOL) instrument version 5. The data were analyzed by using the SPSS 21version. Descriptively: frequency, mean, standard deviation, and inferential statistics: t-Test and one-way analysis of variance (ANOVA), and multiple linear regression analysis were computed. Result The majority of respondents 154 (67.0%) were female. The age of the participants ranges from 20 to 65 (32.06 + 7.45) years. The mean (SD) scores for the dimensions of compassion satisfaction, burnout, and compassion fatigue were 34.41 (6.74), 27.70 (4.24), and 35.83 (7.78) respectively. Neuroticism personality trait had positivity related to compassion fatigue (P = 0.001). Nurses who received low monthly income had significantly lower scores for compassion fatigue (P = 0.002). We found friend support, openness, sex, and agreeableness explained 32.7% (p < 0.024) of the variances in compassion

263777256

satisfaction. Conclusion In general the study found high compassion fatigue and low compassion satisfaction. Further, having low income and neuroticism personality were related to compassion fatigue, while agreeableness, consciousness, and openness personality were related to compassion satisfaction. Therefore, attention should be given to nurses working in cancer centers to ensure positive energy. Introduction Compassionate care for cancer patients may well be a source of both personal fulfillment and intellectual stimulation for healthcare professionals.However, the negative consequences of compassion have been described in terms of compassion fatigue and burnout may threaten professional life in the healthcare system.In particular, professionals working in cancer care are exposed to exact a toll on their physical and emotional health [1].Compassion fatigue reduced the practitioner's capacity to be empathic or bear the suffering of clients [2][3][4][5][6].On the other hand, compassion satisfaction reflects the rewards of caring for others and counterbalances the risks of compassion fatigue [6][7][8][9][10][11][12][13].The finding of the meta-analysis confirmed that there are levels of risk of suffering from burnout and compassion fatigue among nursing professionals.In that study more women and nurses with more years of experience, with nurses from oncology units were affected, having one of the highest levels of burnout and compassion fatigue [14]. The findings in China showed medium levels of compassion satisfaction, burnout, and compassion fatigue among oncology healthcare professionals, reaching rates of 78.34%, 63.50%, and 75.96%, respectively [15].According to a meta-analytic estimation, the prevalence revealed 19% for low compassion satisfaction, 56% for medium and high burnout, and 60% for medium and high compassion fatigue [14].The study

263777256

in Uganda indicated that close to 50% of the nurses experienced compassion fatigue [6].A study conducted in Ethiopia showed that 44.4% of nurses had experienced burnout [9]. The factors that contribute to the negative consequence of compassion are demographic data including age, sex, educational status, area of work, years of experience, social support, experience in oncology care areas, position in the work area, and number of patient assignments [4,6,14].Studies revealed older participants presented higher scores of compassion satisfaction, and younger nurses, women nurses having less job experience, and nurses without leisure activities showed higher means of compassion fatigue [6,8]. In preventing the negative effect of compassionate care, a study identified that nursing staff with self-compassion have a better chance of managing the stresses of their work and care environment [11].Another study highlighted the importance of recognizing individual signs and symptoms of stress, compassion fatigue, and burnout, and then normalizing stress, compassion fatigue, and burnout for health professionals.Moreover, a study mentioned that professionals have to learn how to manage their stress [16]. Despite there were tried to combat the negative consequences of compassionate care, the problem is not reduced among healthcare providers.Moreover, there is a scarcity of studies in Ethiopia.Therefore, this finding may provide insights for hospital leaders, health programmers, and oncology healthcare providers to develop various strategies to improve the professional quality of life among nursing professionals in oncology centers. Therefore; this study aims to quantify the prevalence of compassion satisfaction, burnout, compassion fatigue, and associated factors among nurses working in cancer centers of selected hospitals

263777256

in Ethiopia. Study setting, study design, and period The study was conducted in 5 randomly selected public hospitals that provide cancer treatment in Ethiopia.The selected hospitals were Tikur Anbessa Specialized Hospital (TASH), St Paul's Hospital, Zewditu Hospital in Addis Ababa City, Ayder comprehensive specialized hospital located in Mekelle, Tigray regional State, and Jimma University Medical Center (JUMC) located in Jimma town. A global cancer observer estimated 77, 352 new cancer cases annually in Ethiopia [17].There are 44 medical oncologists and 250 nurses in the oncology centers of the country.The lack of capacity of health facilities outside Addis Ababa has resulted in an enormous load of patients from the rest of the country gravitating towards Tikur Anbessa Specialized Hospital in which there are only 19 beds.Tikur Anbessa Specialized Hospital is the only referral oncology center in the country; about 80% of reported cases of cancer are diagnosed at advanced stages [18]. Institutional-based Cross-sectional study design was used from May 2020 to April 2020 in selected hospitals. Study participants The authors considered eight functional oncology centers in Ethiopia including Ayder Comprehensive Specialized Hospital, Gondar Comprehensive Specialized Teaching Hospital, Felege Hiwot Referral Hospital, Hawassa Referral Hospital, Jimma University Medical Center (JUMC), St. Paul's Hospital, Tikur Anbessa Specialized Hospital and Zewditu Hospital.Among eight hospitals five hospitals were selected using simple random sampling.Tikur Anbessa Specialized Hospital, Ayder Comprehensive Specialized Hospital, St. Paul's Hospital, Jimma University Medical Center.(JUMC).and Zewditu Hospital were the selected areas of the study.Information regarding the study participants was obtained from the human resource management of each institution The

263777256

study participants were nurses working in the cancer treatment centers of the selected hospitals excluding those who are critically sick, thus, unable to fill out the questionnaire and have less than six months of working experience in the cancer treatment unit.Hence, few nurses were practiced in the area of the study, all the 250 nurses who work in cancer centers were selected using the census method. Operational/terms definition Empathy: Empathy is the ability to understand and share other people's emotions and feelings [19,20].Compassion: Compassion is a virtuous and intentional response to knowing a person discerns their needs and ameliorates their suffering through relational understanding and action [19].Professional Quality of Life: Professional quality of life is the internal sense that helpers feel about their work, including both positive and negative aspects [19].Personality traits: Personality traits are an individual's behavior toward others that reflect people's characteristic patterns of thoughts, feelings, attitude, and mindset to make his personality [19]. • Compassion Satisfaction, Compassion Fatigue, and Burnout Independent variables Measurement tools The data was collected using a questionnaire that was developed in the English language and adapted from previously conducted studies in Ethiopia, Nepal, and Spain, [9,20,21].The questionnaire comprised structured questions based on the Professional Quality of Life Scale (PROQOL) instrument version 5. To measure compassion satisfaction for positive feelings and negative feelings of compassion fatigue and burn out we used a self-reported measuring tool.The previously developed ProQOL self-report measuring tool assessed compassion fatigue, compassion satisfaction, and burnout based on how frequently a person has experienced certain antecedents (e.g., "I

263777256

am happy" or "I feel trapped by my job as a helper") within the past 30 days.The ProQOL is comprised of 30 items (10 items for each subscale) that are reflective of the three subscales' content.Items are rated on a five-point Likert-type scale, ranging from 1 (never) to 5 (very often).Several items on the Burnout Subscale are reverse-scored [19]. To test the Pro-QOL 5's reliability within the study sample, Cronbach's Alpha was computed for each of the measured scales.All three scales indicated a high level of internal consistency.Compassion satisfaction (α = 0.831), burnout (α = 0.810), and compassion fatigue (α = 0.781) had high levels of internal consistency. The questionnaires focused on the nurses' level of compassion satisfaction, burnout, and compassion fatigue, and associated factors.The positive and negative ways of compassionate care that can affect the nurse in health care provision were considered [19,22,23]. The data collection instruments include four major parts: Part I: This section includes questions related to socio-demographic information of the study subjects such as sex, age, educational status, monthly income, and work experience.Part II: It includes questions on personality scale questions focused on the individual levels of stress, feeling, and interests.Part III: This section includes questions related to the social support of family and friends Part IV: The questions ask about the nurse's work situation that affects them in both positive and negative, as a helping profession. Data collection procedure Approval from the Ethics Committee of the Addis Ababa University, College of Health Sciences, School of Nursing and Midwifery was obtained before

263777256

data collection.The researchers then contacted the nursing department directors at each hospital and explained the purpose and meaning of this study.Once permission was granted, the head nurses of select departments were invited to serve as research assistants to explain the study's purpose to the participants.After oral informed consent forms were obtained, the researchers distributed the fouritem instruments to the participant.The participants completed the instruments anonymously.All data were numerically coded and accessible only to the researchers to protect confidentiality. Data quality assurance A structured questionnaire was used to collect data after pre-testing the instrument in Armed Force Hospital on 5% of the sample size.The pre-test was conducted two weeks before the actual data collection period and a clear flow in asking the negative and positive questions was maintained.The pretested data were not included in the main study.Complete data collecting instrument preparation takes eight weeks.The data collectors were BSc nurse students who were taken from the same University Hospital in each study area.The training was provided for supervisors and data collectors on the data collection instruments.Then the gap between the method and materials was identified and the appropriate correction was made. Data analysis In this study, data were analyzed by using descriptive statistics including, frequency mean, median, range, standard deviation, and inferential statistics.The normal distribution of the values was tested using the Kolmogorov-Smirnov test.The differences in compassion satisfaction, compassion fatigue, and burnout among participants with demographic and work-related characteristics were tested using the independent t-test and one-way analysis of Variance (ANOVA).Simple regression was performed to assess the relationship

263777256

between the continuous variables and three dependent ones.Multiple linear regression was computed [20].Data analysis was done using SPSS 21.0 statistical program. Socio-demographic and work-related characteristics of the respondents Table 1 shows that out of the total 250 study samples, 230 (92%) individuals responded to our study the majority of respondents accounting for 154 (67.0%) were female.The age of the participants ranged from 20 to 65 (32.06 + 7.45) years.The participants practiced for service age ranged from 1 to 35 (7.75±6.06)years.The oncology unit nursing service experience age ranged from 1 to 16 (4.0±2.91) years (Table 1). Univariate analyses of the factors associated with compassion satisfaction, burnout, and compassion fatigue Table 2 shows the factors associated with compassion satisfaction, burnout, and compassion fatigue.In this study, t-Tests revealed that male nurses had lower compassion satisfaction than female nurses (p = 0.007).Nurses who received low monthly income had significantly lower scores for compassion fatigue (p = 0.002).Nurses' educational status was significantly associated with compassion fatigue (P = 0.002).Nurses who served a few years in the oncology unit had lower compassion fatigue (P = 0.001).Nurses' duty shift was significantly associated with compassion fatigue (P = 0.021).burnout, and compassion fatigue respectively.The personality trait variables in particular agreeableness personality, consciousness, and openness personality had a significant association with compassion satisfaction (all p < 0.035).Neuroticism personality trait had positivity associated with compassion fatigue (all p = 0.002).Compassion fatigue had significant linear relationships with all of the variables of social support (all p < 0.006).In terms of social support variables, friend support had positively related

263777256

to compassion fatigue (p = 0.006).Analyses of the residuals identified that they were normally distributed. Multiple linear regressions statistical analysis of predictors According to Discussion This study was conducted to measure levels of compassion satisfaction, burnout, and compassion fatigue among nurses who work in oncology care units by using the Professional Quality of Life Scale (ProQOL R-V) at five hospitals in Ethiopia [19,23].Registered nurses (N=230) from oncology care units completed the demographic and Professional Quality of Life Scale, Version 5 (Pro-QOL5) questionnaire.In this study, the greatest number of the study participants had a BSc degree (89.1%) educational level.More than half of the respondents were female (67.0%).The study revealed that the mean score of compassion satisfaction, burnout, and compassion fatigue were 34.41, 27.70, and 35.83 respectively.The compassion satisfaction and burnout result is in agreement with the study conducted in Kenya which had a 38.47 mean score of compassion satisfaction and 24.20 burnout, and compassion fatigue in the study (26.33) [4]. This study revealed that male nurses had lower compassion satisfaction than female nurses (p = 0.007).This finding is consistent with the results of another study conducted in Portugal [8].It suggests that males in our setting hardly manage their stress.Friend support, openness, sex, and agreeableness explained 32.7% of the variance in compassion satisfaction.The possible explanation might be attributable to an increased personal stressmanaging ability and a good social support network, Thus, good social support could alleviate the stressful effects of patient care and serve against burnout and compassion fatigue [12].We found that the personality trait variables in particular

263777256

agreeableness, consciousness, and openness personality contributed to compassion satisfaction (p < 0.035).This finding is consistent with the study conducted in China [23].It suggests that open nurses may engage in more activities that enhance their satisfaction with caring activities.Similarly, conscientious nurses are prudent, and hardworking and set high standards for themselves. In this study, family support accounted for 4.7% of the variables in burnout.This finding is in agreement with the study conducted in China that indicates social support acted as a protective predictor of burnout [23]. The six variables of neuroticism, oncology service, educational status, family support, friend support, and significant other support explained 26.3% of the variance in compassion fatigue.This finding is in agreement with a study conducted in China [23].The possible expansion could be neurotic nurses may have a greater inability to control their emotions when faced with negative events, putting them at higher risks of compassion fatigue. Limitations of the study The limitation of the study includes: first, it could be noted from the finding of this study was a small sample size due to the lower number of oncology nurses to draw a random sample.Second, the study also used self-report instruments with the possibility of recall bias because the reliability of collected data can be affected by the respondents' interests and attitudes.Third, this study used a single tool that did not measure the quality of nurses' profession comprehensively.Finally, the study also measured compassion satisfaction; burnout, and compassion fatigue at a single point in time, and the responses could change over time depending on

263777256

respondents' personal and employment situations. Conclusion In general the study finds out that oncology nurses have high compassion fatigue and low compassion satisfaction.Compassionately assisting patients over a long period makes oncology nurses prone to suffer from compassion fatigue.This finding suggests that nurses may be lacking sufficient skills in coping with the traumatic experiences of their patients.Being male, having low income, having longer work experience, and neuroticism personality were related to compassion fatigue, while agreeableness, consciousness, and openness personality were related to compassion satisfaction.It could be concluded that oncology nurses in Ethiopia are under a great deal of hassle and work burdens which leave them defenseless against the winds of Compassion Fatigue and Burnout Out with less Compassion Satisfaction.Therefore, attention should be given to oncology nurses to ensure positive energy, emotional intelligence, and emotional strength. Recommendations This result can be used by nursing leadership and other stakeholders to create an enabling environment for nurses to promote compassion satisfaction and avoid Compassion Fatigue and Burnout in the nursing workforce. Nursing educators should promote relevant on-duty training for oncology nurses and raise their awareness of both the possible negative influences of working with cancer patients and the potential for compassion satisfaction. Further studies should be conducted in different specialties nurses to promote the generalizability of findings and explore other potential predictors. Table 2 Univariate analysis of compassion satisfaction, burnout, and compassion fatigue constructs (N = 230) Table 3 Simple regression for social support, personality traits, and the Compassion satisfaction, Burnout, and Compassion fatigue constructs (N = 230) Table 4

263777256

Multiple linear regression statistical analysis of compassion satisfaction, burnout, and Compassion fatigue among oncology nurses in selected Ethiopian hospitals, (N = 230)

263777256

Satellite DNA-Like Elements Associated With Genes Within Euchromatin of the Beetle Tribolium castaneum In the red flour beetle Tribolium castaneum the major TCAST satellite DNA accounts for 35% of the genome and encompasses the pericentromeric regions of all chromosomes. Because of the presence of transcriptional regulatory elements and transcriptional activity in these sequences, TCAST satellite DNAs also have been proposed to be modulators of gene expression within euchromatin. Here, we analyze the distribution of TCAST homologous repeats in T. castaneum euchromatin and study their association with genes as well as their potential gene regulatory role. We identified 68 arrays composed of TCAST-like elements distributed on all chromosomes. Based on sequence characteristics the arrays were composed of two types of TCAST-like elements. The first type consists of TCAST satellite-like elements in the form of partial monomers or tandemly arranged monomers, up to tetramers, whereas the second type consists of TCAST-like elements embedded with a complex unit that resembles a DNA transposon. TCAST-like elements were also found in the 5′ untranslated region (UTR) of the CR1-3_TCa retrotransposon, and therefore retrotransposition may have contributed to their dispersion throughout the genome. No significant difference in the homogenization of dispersed TCAST-like elements was found either at the level of local arrays or chromosomes nor among different chromosomes. Of 68 TCAST-like elements, 29 were located within introns, with the remaining elements flanked by genes within a 262 to 404,270 nt range. TCAST-like elements are statistically overrepresented near genes with immunoglobulin-like domains attesting to their nonrandom distribution and a possible gene regulatory role.

9737941

repetitive DNA satellite DNA gene regulation transposon immunoglobulinlike genes Based on the hypothesis of Britten and Davidson (1971), repetitive elements can be a source of regulatory sequences and act to distribute regulatory elements throughout the genome. In particular, mobile transposable elements (TEs) are predicted to be a source of noncoding material that allows for the emergence of genetic novelty and influences evolution of gene regulatory networks (Feschotte 2008). Recently it has been shown that at least 5.5% of conserved noncoding elements unique to mammals originate from mobile elements and are preferentially located close to genes involved in development and transcription regulation (Lowe et al. 2007). The complete sequence conservation, wide evolutionary distribution, and presence of functional elements such as promoters and transcription factor binding sites within some satellite DNA sequences has led to the assumption that in addition to participating in centromere formation, they might also act as cis-regulatory elements of gene expression (Ugarkovi c 2005). To perform potential regulatory functions, satellite DNA elements are predicted to be preferentially distributed in euchromatic portion of the genomes in the vicinity of genes. Whole-genome sequencing projects enable the presence and distribution of satellite DNA repeats in the euchromatic portion of the genome to be determined. The analysis of satellite DNA-like elements dispersed within euchromatin, and their comparison with homologous elements present within heterochromatin, also may reveal insights into the origin of satellite DNAs and their subsequent evolution (Kuhn et al. 2012). Satellite DNAs are major building elements of pericentromeric and centromeric heterochromatin in many eukaryotic species, and in

9737941

certain species they account for the majority of genomic DNA, as in beetles from the coleopteran family Tenebrionidae (Ugarkovi c and Plohl 2002). In the red flour beetle Tribolium castaneum, pericentromeric heterochromatin comprises approximately 40% of the genome, and TCAST satellite DNA has previously been characterized as the major satellite that encompasses centromeric as well as pericentromeric regions of all 20 chromosomes (Ugarkovi c et al. 1996). TCAST satellite is composed of two subfamilies, Tcast1a and Tcast1b, which together comprise 35% of the whole genome. Tcast1a and Tcast1b have an average homology of 79% and are a similar size at 362 bp and 377 bp, respectively, but they are characterized by a divergent, subfamily specific region of approximately 100 bp . The genome sequencing project of T. castaneum has recently been completed (Richards et al. 2008). Sequencing involved the euchromatic portion of the genome, with .20% of the genome, corresponding to heterochromatic regions, excluded due to technical difficulties. In this article, we searched for the presence of TCAST satellitehomologous elements within the assembled T. castaneum genome by using a comprehensive computational analysis. By searching the sequenced T. castaneum genome, we found 68 TCAST satellite DNA arrays within the euchromatin of all chromosomes. They were mapped to 59 or 39 ends, as well as within introns, of more than 100 protein-coding genes. Based on sequence characteristics, dispersed TCAST-like elements were classified into two groups. The first group includes partial TCAST satellite monomers or short arrays of tandemly arranged monomers up to tetramers. The second group contains TCAST-like

9737941

element embedded within complex repeat units that contain two hallmarks of DNA transposons, terminal inverted repeats and target-size duplications. The evolutionary relationship and possible modes of dispersion of the two types of dispersed TCAST-like sequences are discussed. In addition, we examined the sequence divergence, phylogenetic relationship, and chromosomal distribution of the elements. Annotation, characterization, and classification of genes within the region of TCAST-like elements are reported, with the preferential localization of TCAST-like elements near specific groups of genes identified. Our results demonstrate for the first time, the enrichment of satellite DNA-like elements in the vicinity of genes with immunoglobulin-like domains and suggest their possible gene-regulatory role. MATERIALS AND METHODS BLASTN version 2.2.22+ was used to screen the NCBI refseq_genomic database of T. castaneum. All scaffolds that have not been mapped to linkage groups were also screened. The program was optimized to search for highly similar sequences (megablast) to the query sequence [TCAST consensus sequence (Ugarkovi c et al. 1996)]. Genes flanking TCAST-homologous elements were found automatically by NCBI blast. Sequences corresponding to hits, as well as their flanking regions, were analyzed by dot plot (http://www.vivo.colostate.edu/ molkit/dnadot/), using standard parameters (window size 9, mismatch limit 0), or more relaxed conditions (window size 11, mismatch limit 1), to determine the exact start and end site of specific TCAST-like elements. The TCAST transposon-like elements were analyzed in detail for the presence of hallmarks such as terminal inverted repeats (TIRs) and target-site duplications with the aid of the Gene Jockey sequence analysis program (for Apple Macintosh). Secondary structures were determined

9737941

using the default parameters of the MFOLD program available online [http://mfold.rna.albany.edu/?q=mfold (Zuker 2003)]. AT content was analyzed using BioEdit Sequence Alignment Editor (Hall 1999). Repbase, a reference database of eukaryotic repetitive DNA, was screened using WU-BLAST (Kohany et al. 2006). Sequence alignment was performed using MUSCLE algorithm (Edgar 2004) combined with manual adjustment. All sequences were included in the alignment, with the exception of the ones that did not at least partially overlap with other sequences. Gblocks was used to eliminate poorly aligned positions and divergent regions of the alignments (Talavera and Castresana 2007). Alignments (original fasta files) are available upon request. jModelTest 0.1.1 software (Posada 2008) was used to infer best-fit models of DNA evolution-TPM3uf+G for transposon-like and A type elements and TPM1uf for B type elements. Maximum likelihood (ML) trees were estimated with the PhyML 3.0 software (Guindon and Gascuel 2003) using best-fit models. Markov chain Monte Carlo Bayesian searches were performed in MrBayes v. 3.1.2. (Huelsenbeck and Ronquist 2001) under the best-fit models (two simultaneous runs, each with four chains; 3 · 10 6 generations; sampling frequency one in every 100 generations; majority rule consensus trees constructed based on trees sampled after burn-in). Branch support was evaluated by bootstrap analysis (1000 replicates) in ML and by posterior probabilities in Bayesian analyses. Pairwise sequence diversity (uncorrected P) was calculated using the MEGA 5.05 software (Tamura et al. 2011). T. castaneum gene homologs in Drosophila melanogaster were searched using the OrthoDB Phylogenomic database. Each gene has OrthoDB identificator, with Uniprot data linked to OrthoDB (Waterhouse

9737941

et al. 2011). To find sets of biological annotations that frequently appear together and are significantly enriched in a set of genes located near TCAST-like elements, program GeneCodis 2.0 available online (http://genecodis.dacya.ucm.es/) was used. GeneCodis generates statistical rank scores for single annotations and their combinations. To find all the possible combinations of annotations, Gen-eCodis uses the apriori algorithm introduced by Agrawal et al. 1993. Once the annotations were extracted, a statistical analysis based on the hypergeometric distribution or the x 2 test of independence was executed to calculate the statistical significance (P values) for each individual annotation or co-annotations. Two-tailed hypergeometric test with Bonferroni correction (alpha = 0.025) was used to analyze the distribution of TCAST-like elements among T. castaneum chromosomes. In each chromosome the frequency of TCAST-like elements was compared with the frequency in the complete sample and the significance of deviations was calculated. Identification of dispersed TCAST-like elements Using the consensus sequence of TCAST satellite DNA (Ugarkovi c et al. 1996) as a query sequence, we screened the NCBI refseq_ genomic database of T. castaneum with the alignment program BLASTN version 2.2.22+. The program was optimized to search for highly similar sequences (megablast) and blast hits on the query sequence were analyzed individually. Alignments were mapped regarding start and end site, chromosome number, and total length. When the distance between two alignments on the same chromosome was short, the genomic sequence was further analyzed by dot plot to identify any potential continuity between the two alignments. Only genomic sequences with at least 140 nt

9737941

(40% of TCAST monomer length) of continuous sequence and .80% identity to the TCAST consensus sequence were considered for further analysis. The total number of dispersed TCAST-like elements was 68, with 36 elements flanked by genes at both 59 and 39 ends, 3 elements flanked by a single gene either at 59 or 39 end (sequences no. 36, 39, 50), and the 29 elements positioned within introns (Table 1). Except 68 TCAST-like elements associated with genes, no other dispersed TCAST-like elements were found within the assembled T. castaneum genome. Analysis of scaffolds that have not been mapped to linkage groups revealed the presence of an additional 41 TCAST-like elements, but because they were not mapped to T. castaneum genome and could possibly derive from heterochromatin, we did not consider them for further analysis. There were only three cases in which two different TCAST-like elements were associated with the same gene: gene D6X2C4 contains TCAST-like sequences no. 6 and 13 within introns, gene D6X2U7 is flanked at 59 and 39 end by sequences no. 5 and 7, respectively, whereas gene D6WB29 is located at 39 end of the sequence no. 53 and has sequence no. 52 within an intron. All other TCAST-like elements were positioned near or within different genes. Thus in total, there were 101 genes found in the vicinity of TCAST-like elements. Characteristics of the genes associated with TCAST-like elements, including gene identity number, gene name and chromosomal location, position relative to the associated TCAST-like element, and distances between TCAST-like elements and genes, are shown

9737941

in Table 1 Distances between TCAST-like elements and genes range from 262 nt (gene positioned at 39 site of the sequence no. 36), to a maximal distance of 404,270 nt (gene positioned at 59 site of the sequence no. 5). Characteristics of TCAST-like elements TCAST satellite-like elements: Sequence analysis of the 68 TCASTlike elements identified within the vicinity of genes enabled their classification into two groups. The first group contains partial TCAST satellite monomers or tandemly arranged elements, either complete or partial dimers, trimers, or tetramers ( Table 1). The minimal size of satellite repeat was 203 nt (0.6 of complete TCAST monomer; sequence no. 15), whereas the maximal size was 1440 nt (four complete TCAST monomers; sequence no. 43; Table 1). In many sequences, two subtypes of TCAST satellite monomers were mutually interspersed: Tcast1a and Tcast1b. Tcast1b corresponds to the TCAST satellite consensus that was used as a query sequence (Ugarkovi c et al. 1996), and Tcast1a corresponds to the TCAST subfamily described in Feliciello et al. 2011. Tcast1a and Tcast1b have an average homology of 79% and are of similar sizes at 362 bp and 377 bp respectively, but are characterized by a divergent, subfamily specific region of approximately 100 bp ). There were 34 TCAST satellitelike elements found within or in the region of 53 genes. Lengths of TCAST satellite-like elements (Table 1), their exact start and end sites within genomic sequence and composition (supporting information, Table S1) are provided. To see whether there is any clustering of sequences of TCAST satellite-like elements

9737941

due to the difference in the homogenization at the level of local array, chromosome, or among different chromosomes, sequence alignment and phylogenetic analysis were performed. Tcast1a and Tcast1b subunits were extracted from TCAST satellitelike sequences and analyzed separately. Alignment was performed on 24 Tcast1a subunits, ranging in size from 136 and 377 bp (File S1). The average pairwise distances between Tcast1a subunits of TCAST satellite-like sequences was 5.8%. Alignment adjustment using Gblocks, which eliminates poorly aligned positions and divergent regions, resulted in few changes; therefore, the original, unadjusted alignment was used for the construction of phylogenetic trees. Because the sequences differ in lengths and comprise regions of divergent variability, methods that take into account specific models of DNA evolution were considered as the most suitable for the construction of phylogenetic trees, ML and Bayesian (Markov chain Monte Carlo). The ML tree showed weak resolution with no significant support for clustering of sequences derived from the same satellite-like array or from the same chromosome. Similarly, the Bayesian tree demonstrated no significant sequence clustering ( Figure 1A). Alignment of 28 Tcast1b subunits, ranging from 159 bp to 363 bp (File S2), was also not significantly affected by adjustment with Gblocks; therefore, the unadjusted alignment was used for the construction of phylogenetic trees ( Figure 1B). The average pairwise divergence between Tcast1b subunits, of TCAST satellite-like sequences, was 6.7%. With the ML phylogenetic tree, four groups composed of two or three sequences, were resolved by relatively low bootstrap values. However, the majority of Tcast1b subunit sequences remained unresolved. There

9737941

was no clustering of subunits derived from the same array or the same chromosome ( Figure 1B). Bayesian tree analysis produced one significantly supported cluster composed of 10 sequences derived from 7 chromosomes ( Figure 1B). TCAST transposon-like elements: The second group of TCAST-like repeats is represented by a complex element that contains an almost complete TCAST (or Tcast1b) monomer, and a TCAST monomer segment of approximately 121 bp in an inverted orientation. These two TCAST segments are separated by a nonsatellite sequence of approximately 306 bp. Both TCAST segments are part of TIRs that are approximately 269 bp long ( Figure 2). As a result of the long TIRs, these elements are likely to form stable hairpin secondary structures and therefore resemble transposons. The nonsatellite part of sequence, common for all TCAST transposon-like elements, is unique in that it does not exhibit significant homology to any other sequence within the T. castaneum genome. There were 34 TCAST transposon-like elements found within or in the vicinity of 50 genes. Their lengths (Table 1) and exact start and end sites within genomic sequence (Table S1) are provided. Sequence analysis of TCAST transposon-like elements determined that 13 of them were . 1000 bp, with a maximal size of 1181 bp (Table 1). The remaining TCAST transposon-like elements were shorter, with a minimal size of 314 bp (sequence no. 27), and usually lacking part of, or one or both, TIRs. Conserved TIRs are necessary for transposition, and if they are absent, truncated, or mutated so that the transposase cannot

9737941

interact with the transposon sequence, the transposon cannot be mobilized and therefore represents a molecular fossil of a once active transposon (Capy et al. 1998). Despite mutations and partial truncations of TIRs within the TCAST transposon-like elements, and likely because of the length of the TIRs, most of the elements still preserve a stable secondary structure and could potentially remain mobile. Some TCAST transposon-like elements .1000 bp have a 3-bp duplication at the site of insertion in the form of ACT. One TCAST transposon-like element (sequence no. 39) is inserted into another repetitive DNA, indicated as Tcast2, which had been previously identified bioinformatically (Wang et al. 2008). Sequence analysis of this transposon-like element confirms the continuity of Tcast2 from the duplication site "ACT." Typically, the size of target-site duplication is a hallmark of different superfamilies of eukaryotic DNA transposons, with mariner/Tc1, the only superfamily whose members are characterized by either 2-or 3-bp target-site duplication (Capy et al. 1998;Kapitonov and Jurka 2003;Feschotte and Pritham 2007). There are three open reading frames (ORFs) within TCAST transposon-like sequences, but the resulting putative proteins are very short and do not share similarity with any other proteins ( Figure 2). The elements therefore do not code for transposases and are considered nonautonomous. Using the whole TCAST transposon-like elements as a query n sequence, we searched the T. castaneum Gen Bank database for "fullsized" homologous elements that could potentially code for transposases and be considered autonomous. The search identified an element, named TR 1.9, with a 925-bp sequence inserted within a unique

9737941

sequence of the TCAST transposon-like elements (Figure 2). This 925-bp sequence contains an ORF of 206 amino acids and a conserved domain belonging to the Transposase 1 superfamily, which also includes the mariner transposase. DNA transposons of the mariner/Tc1 superfamily Mariner-1_TCa and Mariner-2_TCa, were identified within the T. castaneum genome (Jurka 2009a(Jurka , 2009b. Using BLASTP and the translated sequence from the 925 bp ORF as a query sequence, we identified hits with a partial homology to a Mariner-2_TCa transposase and to a mariner-like element transposase present in two other insects, the beetle Agrilus planipennis (emerald ash borer) and Chrysoperla plorabunda (green lacewing; Neuroptera), but not to Mariner-1_ TCa transposase. To test whether there is any chromosome-specific sequence clustering of TCAST transposon-like sequences that could suggest difference in homogenization within chromosome and among different chromosomes, the alignment and subsequent phylogenetic analysis of TCAST transposon-like sequences was performed. Because TCAST transposon-like elements differ significantly in size (31421181 nt), the alignment and phylogenetic analyses was performed on 25 elements that mutually overlap in their sequences, whereas the other nine TCAST transposon-like elements were excluded from the analysis due to the very low overlapping with other elements. Alignment was additionally adjusted using Gblocks (File S3). The average pairwise divergence among TCAST transposon-like sequences was 12.7%. ML and Bayesian methods gave similar tree topologies ( Figure 1C). The ML tree showed very weak resolution of TCAST transposon-like sequences and a general absence of subgroups with specific sequence characteristics ( Figure 1C). Only two clusters were formed whereas, using the Bayesian

9737941

tree, we identified three well-supported groups; two of them were as for ML tree ( Figure 1C). Distribution of TCAST-like elements on T. castaneum chromosomes TCAST-like elements found in the vicinity of genes were distributed on all 10 T. castaneum chromosomes ( Table 1). Positions of constitutive heterochromatin and euchromatin were assigned on the haploid set of T. castaneum chromosomes, based on C-banding data (Stuart and Mocelin 1995) and Tribolium castaneum 3.0 Assembly data (Figure 3). Within euchromatic segments, the position of each TCASTlike element is specifically indicated (Figure 3) based on the position within the genomic sequence (Table S1). TCAST-like elements were dispersed on both arms of chromosomes 3, 5, 9, and 7, whereas on other chromosomes they were located on a single arm (Figure 3). The number of TCAST-like elements ranged from 2 on chromosome 1(X) to 17 on chromosomes 3 and 9. To detect whether TCAST-like elements were distributed randomly among the T. castaneum chromosomes or whether there was a significant over or underrepresentation of the elements on some chromosomes we performed hypergeometric distribution analysis test. The analysis revealed no statistically significant deviation in the number of TCAST-like elements among the chromosomes ( Figure S1), pointing to their random distribution. To determine whether there was a target preference for the insertion of TCAST-like elements, for example high AT content or another sequence characteristic, we analyzed the AT content within 100 bp of the flanking regions for each TCAST -like element, from both 59 and 39 sites ( Figure S2 and Figure S3). The

9737941

average AT content n A list of genes with gene identity numbers, gene name, chromosomal location, position, and distance relative to the associated TCAST-like element, as well as a list of TCAST-like elements, their types (satellite or transposon-like), total length in bp, and copy number of satellite repeats within an array are shown. of the flanking regions for both TCAST satellite-like elements and TCAST transposon-like elements did not differ significantly from the average AT content of the whole T. castaneum genome or from the AT content of randomly selected intergenic regions and introns. Thus, this finding suggests that with regard to AT content, there is no target preference for the insertion of TCAST-like elements. Furthermore, alignment and comparison of all flanking sequences of TCAST-like elements did not identify any common sequence motifs. Genes in the vicinity of TCAST-like elements Uniprot gene numbers were used as identifiers of genes located in the vicinity of TCAST-like elements (gene names shown in Table 1). Uniprot gene numbers for homologous genes found in Drosophila melanogaster are also indicated (Table 1). Detailed description of the genes, including molecular function of their protein products, biological processes in which these proteins are involved, and their cellular localization (cellular component), are shown (Table S1). Each identified gene is assigned to a particular TCAST-like element within its vicinity, and the precise position of TCAST-like elements in genomic sequence (start and end site) is indicated (Table S1). Functional analysis revealed that 17 of 101 genes correspond to putative uncharacterized proteins, whereas the remaining genes are involved

9737941

in different molecular functions and diverse biological processes. Among the proteins, a proportion is characterized by ATP binding activity (13 proteins) and involvement in protein phosphorylation and /or signal transduction (9 proteins; Table S1). To determine whether TCAST-like elements are distributed randomly relative to genes or whether they are overrepresented near specific groups of genes, we used GeneCodis 2.0 to provide a statistical representation of the genes associated with TCAST-like elements. Because many genes are still not annotated in T. castaneum and furthermore T. castaneum genomic data are not included in Gene-Codis, we used gene numbers for orthologous genes from D. mela-nogaster for the analysis and compared them with the whole set of 14,869 genes annotated in D. melanogaster. Genecodis analysis revealed that TCAST-like elements are located near nine genes characterized as members of the immunoglobulin protein superfamily. Because there are only 134 immunoglobulin-like genes present within the total set of D. melanogaster genes, random distribution of TCAST-like elements would result in their occurrence near approximately a single immunoglobulin-like gene. The presence of TCAST-like elements in the vicinity of nine immunoglobulin-like genes therefore represents a statistically significant overrepresentation (0.00000427). All nine genes exhibit structural features of immunoglobulin-like, immunoglobulin subtype 1 and immunoglobulin subtype 2 proteins and are associated with the following TCAST transposon-like elements: 25 at the 39end, 28 and 39 at the 59 end, 32 and 40 within introns, and TCAST satellite-like elements: 8 at the 39 end, 19 and 62 at the 59 end, and 41 within intron (Table 1). A minimal distance between

9737941

TCAST-like element and immunoglobulinlike gene was 7165 bp and a maximal 173,881 bp (Table 1). Molecular function of most of immunoglobulin-like genes is unknown, and they are involved in different biological processes such as cell adhesion, protein phosphorylation, and axon guidance (Table S1). Although all nine genes belong to immunoglobulin superfamily, they did not exhibit sequence similarity, which could suggest role of duplication in their evolution and spreading. The position of TCAST-like elements relative to the genes also was not consistent with the possibility that TCAST-like elements duplicated along with the immunoglobulin-like genes. Overrepresentation of TCAST-like elements was also found near genes that exhibit ATP-binding activity and axon guidance properties but with a marginal significance (0.0183374 and 0.00865139). For the rest of genes, no significant overrepresentation of TCAST-like elements was detected. Thus, enrichment of TCAST-like elements in the vicinity of immunoglobulin-like genes potentially implicates a role of TCAST-like elements in the regulation of these genes. Table 1. When a particular sequence is composed of few subrepeats (e.g., Tcast1a or Tcast1b), numbers indicating subrepeats are added (e.g., 43_1, 43_2, 43_3). Numbers in brackets indicate chromosomes on which the corresponding sequences are located. Numbers on branches indicate Bayesian posterior probabilities/ML bootstrap support (above 0.5/50%, respectively). DISCUSSION TEs are classified in several dozen families based on transposition mechanisms and different dynamics properties (Hua- Van et al. 2005). Active TEs encode the enzymes necessary for their transposition, either to move between nonhomologous regions in the genome or to copy themselves to other positions. In many cases, TEs do not produce

9737941

their own enzymes but are able to use those from functional copies or even from other TEs families. Defective and inactive TEs often are amplified in regions of low recombination such as heterochromatin and may form tandemly repeated satellite DNAs. The origin of satellite DNA array from transposon-like elements is reported for many insects such as Drosophila melanogaster (Agudo et al. 1999), Drosophila guanche (Miller et al. 2000), and the beetle Misolampus goudoti (Pons 2004) whereas the retroviral-like features were first observed in the satellite DNA from rodents of the genus Ctenomys (Rossi et al. 1993). Transposons can be inserted into other repetitive sequences such as satellite DNAs, as has been observed for the mariner-like element and MITE element, both inserted into satellite DNA of the ant Messor bouvieri (Palomeque et al. 2006). Searching for repetitive elements homologous to the TCAST repeat within Repbase (http://www. girinst.org/repbase/) revealed that 59 UTR of nonlong terminal repeat retrotransposon CR1-3_TCa (Jurka 2009c) shares a high similarity of 83% with a 444-bp long TCAST sequence composed of 1.2 tandem monomers (Figure 1). Other CR1 subfamilies identified within T. castaneum such as CR1-1_ TCa, CR1-2_TCa, and CR1-4_TCa, published in Repbase, do not share similarity to CR1-3 and do not contain TCAST similar sequence. We propose that CR1-3 was inserted within TCAST satellite array and through recombination has acquired a part of TCAST sequence. Newly acquired TCAST element could act as a promoter because TCAST satellite DNA has an internal promoter for RNA Pol II (Pezer and Ugarkovi c 2012) and becomes a

9737941

new functional 59 UTR. Subsequent retrotransposition of CR1-3_TCa could explain the dispersion of TCAST within the euchromatin (Figure 4). Three CR1-3_TCa elements with TCAST in the 59UTR were identified within scaffolds that have not been mapped to linkage groups. However, truncated fragments with partial homology to CR1-3_TCa retrotransposon can be mapped within T. castaneum genome, some of them in the vicinity of TCAST elements. Such arrangement also indicates the role of CR1-3_TCa in the spreading of TCAST elements. There is also a possibility that TCAST satellite DNA originates from CR1-3 retrotransposon which was, after inactivation, amplified within the heterochromatin region. In the case of TCAST transposon-like elements, part of the satellite sequence is incorporated within TIRs which are characteristic for DNA transposons. The presence of target-site duplications at the sites of insertions of some TCAST transposon-like elements also indicates transposition as a mode of spreading of TCAST Figure 2 Organization of TCAST elements within T. castaneum genome in the form of TCAST transposon-like element, tandem arrays, and CR1-3_TCa retrotransposon. Regions corresponding to TCAST element are shown in red. TCAST transposon-like element contains an almost complete TCAST monomer and a monomer segment of approximately 121 bp in an inverted orientation, whereas CR1-3 retrotransposon contains segment corresponding to 1.2 monomer. Within TCAST transposon-like element terminal inverted repeats (arrows) unique nonsatellite sequence (green), target-site duplication in the form of "ACT," and the insertion point of 925-bp sequence found within TR 1.9, element and coding for the putative transposase are shown. Three short ORFs within TCAST transposon-like element are also

9737941

indicated. Within nonlong terminal repeat retrotransposon CR1-3_TCa regions corresponding to 59UTR and to two ORFs are indicated. Figure 3 Distribution of TCAST-like elements on T. castaneum chromosomes. The karyotype representing the haploid set of T. castaneum chromosomes, and positions of constitutive heterochromatin (dark) and euchromatin (white) are depicted based on C-banding data (Stuart and Mocelin 1995) and T. castaneum 3.0 assembly (http://www.beetlebase.org). TCAST transposon-like elements (blue) and TCAST satellite-like elements (red) are shown. Two TCAST-like elements are represented as separate lines if they are at least 100 kb distant from each other. elements. Parts of satellite DNA elements can be found within some transposons, such as pDv transposon (Evgen'ev et al. 1982;Zelentsova et al. 1986) whose long direct terminal repeats show significant sequence similarity to the pvB370 satellite DNA, located in the centromeric heterochromatin of a number of species of the Drosophila virilis group (Heikkinen et al. 1995). The presence of short stretches of PisTR-A satellite DNA sequences within 39 UTR of Ogre retrotransposons dispersed in the pea (Pisum sativum) genome was reported (Macas et al. 2009). Furthermore, the mobilization of subtelomeric repeats upon excision of the transposable P element from tandemly repeated subtelomeric sequences has been observed ( Thompson-Stewart et al. 1994). Incorporation of part of a TCAST satellite DNA sequence into a (retro)transposable element, and its subsequent mobilization and spreading by (retro)transposition, may explain the distribution of TCAST element in the vicinity of genes within euchromatin. Satellite DNA sequences are prone to undergo recurrent repeat copy number expansion and contraction in divergent lineages as

9737941

well as among populations of the same species (Bosco et al. 2007). This amplification appears to be random and does not correlate with phylogeny of the species Lee et al. 2005;Bulazel et al. 2007). Amplification of a satellite sequence is reported to occur as a result of unequal crossing over or duplicative transposition (Smith 1976;Ma and Jackson 2006). The discovery of human extrachromosomal elements originating from satellite DNA arrays in cultured human cells and different plant species indicates the possible existence of additional amplification mechanisms based on rolling-circle replication (Assum et al. 1993;Navrátilová et al. 2008). It has been proposed that satellite sequences excised from their chromosomal loci via intrastrand recombination could be amplified in this way, followed by reintegration of tandem arrays into the genome (Feliciello et al. 2006). Moreover, it is possible that such a mechanism affected TCAST satellite DNA, and that extrachromosamal circles of TCAST were reintegrated into different genome locations by homologous recombination based on short stretches of sequence similarity between TCAST satellite and target genomic sequence (Figure 4). Integrated TCAST sequences are mainly composed of interspersed elements belonging to two major subfamilies, Tcast1a and Tcast1b, which is a prevalent type of organization in pericentromeric heterochromatin . This finding indicates that the origin of dispersed euchromatic TCAST elements may be duplication of heterochromatin copies. The distribution of TCAST-like elements relative to protein coding genes revealed no specific preference for insertions within introns or at 59 or 39 ends of genes. TCAST-like elements are distributed on all chromosomes with no significant deviation in

9737941

the number among the chromosomes, and phylogenetic analysis did not detect any significant sequence clustering of TCAST-like elements derived from the same chromosome. Dispersed TCAST satellite-like elements produce tandem arrays up to tetramers, but repeats from the same array do not reveal any significant clustering on phylogenetic trees. This finding indicates there is no significant difference in the homogenization of TCAST satellite-like repeats at the level of local arrays or chromosome or among different chromosomes. The average pair-wise sequence divergence (6% for dispersed TCAST satellite-like repeats) is greater than the usual divergence of satellite elements located in heterochromatin of tenebrionid beetles [approximately 2% (Ugarkovi c et al. 1996)]. This difference in homogeneity between repeats located in heterochromatin and euchromatin may be explained by a lower rate of gene conversion affecting dispersed satellite-like elements or by a specific mechanism of DNA repair acting on satellite DNA (Feliciello et al. 2006). TCAST transposon-like elements dispersed among the genes within euchromatin have an even greater average sequence divergence (approximately 12%) and also exhibit no significant chromosome-specific sequence clustering, indicating a similar rate of homogenization within and among the chromosomes. Relatively high Figure 4 Models of spreading of TCAT-like elements based on (A) retrotransposition of CR-3_TCa element. CR1-3_TCa was inserted within TCAST satellite array and through recombination has acquired a part of TCAST sequence, which could act as a promoter and become a new functional 59UTR. Subsequent retrotransposition of CR1-3_TCa could explain the dispersion of TCAST within the euchromatin. (B) Rolling circle replication of TCAST satellite DNA sequences excised from

9737941

their heterochromatin loci via intrastrand recombination, followed by reintegration into different genome locations by homologous recombination. sequence divergence of TCAST transposon-like elements and the significant truncation of the majority of them, indicates that the transposition of these elements did not occur very recently and that these elements could be considered as molecular fossils of the functional TCAST transposon-like elements. Cis-regulatory elements, such as promoters or transcription factor binding sites, are predicted in some satellite DNAs (Pezer et al. 2011). Transcription from promoters for RNA Pol II is also characteristic for pericentromeric satellite DNAs from the beetles Palorus ratzeburgii and Palorus subdepressus (Pezer andUgarkovi c 2008, 2009). Temperature-sensitive transcription of TCAST satellite DNA from an internal RNA Pol II promoter has been demonstrated (Pezer and Ugarkovi c 2012). Based on these findings, it can be proposed that TCAST elements located in the vicinity of genes may function as alternative promoters, and transcripts derived from them may interfere with the expression of neighboring gene. This type of regulation is often observed for retrotransposons positioned immediately 59 of protein genes (Faulkner et al. 2009). In addition, some tissue-specific gene promoters are derived from retrotransposons (Ting et al. 1992;Samuelson et al. 1996). Because of rapid evolutionary turnover, satellite DNA sequences often are restricted to a group of closely related species, or in some instances are species specific. This is the case with TCAST satellite DNA, which is not even detected in the congeneric Tribolium species. If restricted satellite DNAs have regulatory potential, then insertion of these elements in vicinity of

9737941

genes could contribute to the establishment of lineage-specific or species-specific patterns of gene expression. Annotation of genes in proximity to TCAST-like elements demonstrated a statistical overrepresentation of certain groups of genes, for example, those with immunoglobulin-like domains. Recently, in the fish Salvelinus fontinalis, a regulatory role of a 32-bp satellite repeat, located in an intron of the major histocompatibility complex gene (MHIIb), on MHIIb gene expression was demonstrated (Croisetiere et al. 2010). The level of gene expression depends on temperature, as well as the number of satellite repeats, and indicates a role for temperature-sensitive satellite DNA in gene regulation of the adaptive immune response. Further studies are necessary to determine whether TCAST-like elements exhibit a potential regulatory role on nearby genes. The transcriptional potential of satellite DNAs as well as their distribution close to protein-coding genes, as shown in this study, provides strong support, that in addition to transposons, satellite DNAs represent a rich source for the assembly of gene regulatory systems.

9737941

TRANSFORMING THE STATUS AND ROLE OF THE TEACHER AT THE INSTITUTE OF HIGHER EDUCATION IN THE DIGITAL EDUCATIONAL ENVIRONMENT TRANSFORMANDO O STATUS E O PAPEL DO PROFESSOR NO INSTITUTO DE ENSINO SUPERIOR NO AMBIENTE DIGITAL EDUCACIONAL The purpose of this study was to investigate the role of teachers in the modern educational process. To achieve this goal, a large-scale survey of teachers (N=419) and students (N=1372) was conducted in 2022 at Kazan (Volga Region) Federal University (Russia), one of the leading universities in Russia. The research provides insights into the transformation of teaching status and role. The function of transferring theoretical knowledge is gradually giving way to a more student-centered approach, characterized by an increased emphasis on mentoring. The study identifies both traditional and emerging roles performed by higher education teachers in the context of digitalization within the modern information society. INTRODUCTION With the beginning of the IT century, the teacher was no longer the only source of knowledge (Kist, 2010;Margaryan, Littlejohn, & Vojt, 2011). If the traditional task of the teacher was to transfer knowledge and skills to students, the modern teacher at the university is expected to teach students to acquire knowledge, to form independent educational activities, guide the thinking of students on their own professional development, will be fascinated by the idea that knowledge and skills, which were acquired at the university, will be fundamental for further professional activity (Postman, 1992(Postman, , 1995Aleksandrovna, Iurievna, & Olegovna, 2017). Moreover, with the help of new information and communication technologies, it is possible to significantly enhance

259754022

the content of the educational process, to increase the availability and transparency of information resources and means of communication when using them in the educational process (Davidson & Goldberg, 2009;So & Kim, 2018;Maximova, Eflova, & Kulcha, 2018). Along with the development of Internet technologies, all elements of social life change radically: the economy, politics, education, labor relations, and through this the objectives and goals of the learning process, which determines a significant change in the role of the teacher in the institute of higher education (Deursen & Dijk, 2014). The modern institute of higher education requires a change in the roles of teachers and students, there is a transition to a form of cooperation and collaborative activity, which purpose is to develop the abilities that ensure effective professional activity. Thus, the teacher and student become partners in the learning process, and the role of the teacher changes from the giver, examiner, supervisor to helper in acquiring of new knowledge and skills. METHODS The theoretical approach to the analysis of the transformation of the institution of higher education in the context of globalization was formed under the influence of representatives of the information society and sociologists of global culture. In the works of the American anthropologist and researcher of global cultural phenomena Appadurai (1996) and the information society theorist Castells (1989Castells ( , 2004. In their analytical matrix, the authors distinguish three structural characteristics of the cultural and educational process in the context of globalization: transnationalization, self-organization of networks and marketization. The Institute of Higher Education in

259754022

the context of globalization seeks to self-organization and acts as a system of educational services. This vision of the development of the Institution of Higher Education during globalization determines the content of academic discourse. In order to identify the ideas of students and teachers at the Institute of Higher Education about the role of the teacher in the modern educational process, sociological study in the form of a mass survey was conducted in 2022. The survey was attended by 1,372 students of the Kazan (Volga Region) Federal University, representing various gender and age groups of the student community. A survey of KFU professors was also conducted (n=419). Most of the teachers have more than 16 years of teaching experience (46%), 11 to 15 years of teaching -18%, 6 to 10 years of teaching -19%, the rest had less than 5 years of teaching. The results of the research allowed to conduct sociological analysis of data not only in the context of their quantitative assessment, but also qualitative, and to make the following generalizations and conclusions. RESULTS AND DISCUSSION In a survey of teachers on the transformation of their status and role in the context of digitalization, 39% noted the unconditional change in relations between teachers and students associated with the integration of information and communication technologies in the educational process. 42% are not ready to talk about a fundamental change in relations between the participants of the educational process, pointing only to the partial nature of such changes. Only 14% of respondents did not notice the

259754022

changes that came with digitization. First of all, there has been a change in the channels of communication with students, which was noted by the majority of teachers (58%): firstly, the channels of communication are now more and more diverse and secondly, the channels of communication that have been created allow contact with students and teachers at high speed and with less restrictions. It leads to more frequent consultations with teachers on scientific, educational, and other activities, as it no longer requires the expenditure of temporary or other resources, as reported by 37% of respondents. Most of the new channels of communication become possible because of the Internet, which entails some changes in the way students and teachers communicate, due to the perception of virtual reality as something frivolous, far from the educational process. This impression of the Internet was created at the beginning of its appearance and implementation in social institutions. For a long time, scientists in different fields have ignored the study of Internet practices because of their lack of seriousness. The described attitude towards the Internet and digital technologies is a thing of the past, but its departure can be characterized as rather slow, gradual, and therefore unsurprisingly disturbed in the formal style of communication by students. 38% of respondents complain about this, noting the lack of respect and culture of communication from the students. However, the absence of serious attitude to the Internet as a tool of erasing the boundaries between teacher and student confirms the presence among respondents of 17% who

259754022

noted the formalization of relations with the arrival of digital technologies in the educational space. Assessing the merits of digital education, respondents noted faster and easier access to information (51%), easier exchange of information with students (48%), possibility of multiple reproduction (43%), visualization of information (41%) ( Table 1). Table 1. Pluses of the digitalization of education As you can see, the greatest number of advantages is allocated by teachers in the field of the opportunities to work with a huge amount of information, which can be presented not only conveniently, but also more quickly and understandably, which contributes to better learning N Respondents' answers Results, % 1 faster and easier access to information 51% 2 easier exchange of information with students (as a comparison with the past) 48% 3 more flexible organization of the educational process 45% 4 the possibility of multiple playbacks of educational material 43% 5 visualization of information 41% 6 increasing student involvement in learning 26% 7 development of a network model of cooperation between institutes/university faculty 23% 8 Lack of control over the teacher's work 14% of the current discipline, including an ability to replay the information countless times. It is worth mentioning that the students are in complete solidarity with the teachers in this matter: diverse presentation of information and its accessibility are also those factors, which implementation contributes improving the teaching status in the eyes of the students. According to 45% of respondents, more flexible organization of the educational process is also the advantage of digitalization of the educational

259754022

process. Due to the reduction of time and space limitations, a quarter of respondents noted an increase in the inclusion and involvement of students in the educational process. Digital technologies are actively integrated in university life and change educational settings, but they are also a source of digital inequality (table 2). decline in the involvement of students in the learning process 38% 6 the complexity of adapting several disciplines to the digital format 36% 7 lack of teachers' knowledge of digital equipment 32% 8 change in teacher-student relations 28% 9 the complexity of the transition to digital education 25% 10 transition to online education 22% 11 lack of control over the teacher's work 14% face difficulties in the process of transition to digital education and the development of such a new format. Some teachers (22%) see as a minus the gradual replacement of full-time lessons with online education, in which there is a loss of direct interaction with students. As the results of the study showed, a certain hierarchy of teachers' roles was formed through self-identification (figure 1). Figure 1. Current roles of the teacher The leading role of the teacher, according to 54% of respondents, is mentoring, in which the teacher with an individual approach to each student teaches new generations to navigate the information field and develops the personal and professional qualities of students. Half of the respondents also noted the relevance of the role of the speaking teacher as a practitioner. The reproductive method of teaching, the transfer of theoretical knowledge is gradually

259754022

fading, because the new generation of students has access to scientific information beyond classrooms and university libraries (Voogt, et al., 2017). At the same time, the application of this knowledge, working with the available information as valuable skills in modern realities are no longer so easily acquired outside the lectures and seminars. Nevertheless, teachers do not devalue the role of the teacher as a lecturer, as even the format of lecture classes in a digital environment provides a whole range of opportunities for the translation and implementation of the subject. Also, the changing format of the educational process requires coordination, in which the What are the most important roles of teachers in the modern education system? primary role is played by the teacher (37%). The rather conservative role of the teacher as a scientist is also important for teachers. Based on the data, it can be concluded that educators understand the consequences of digitalization, which affects the transformation of teaching status and role. The function of transferring theoretical knowledge is gradually losing ground to a more student-oriented approach, with an increased emphasis on mentoring. Teachers continue to serve as guides between students and valuable knowledge, but the content and methods of the teaching process are changing. It is becoming crucial to teach students how to discern valuable and meaningful information from the vast digital space. Practical application and adaptation of existing knowledge to current realities are gaining importance, necessitating continuous monitoring of new scientific knowledge online and timely updating of teaching materials. Due to the development

259754022

and integration of digital technologies at the university, new communication channels were opened for participants of the educational process, they reduced the formality of communication between teachers and students resulting in violation of the working and personal boundaries of teachers, which respondents assess very negatively. The decline in the prestige of the teaching profession, which was noted by respondents, is largely connected to the digitalization of education. However, despite the perceived weakening of their role in the learning process, respondents tend to be optimistic about the future: by introducing additional courses that could raise awareness of digital equipment of the university, and the level of their mastery, the teaching staff will have more opportunities to facilitate professional activity, increase its convenience, speed, quality, and competence growth through the use of digital media in teaching. The processes that influenced the transformation of the institute of higher education have also changed the status of the teacher. The traditional form of lectures, where the lecturer tells, and students record, goes into the past. Nowadays there is a need to hold lectures using video, presentations, digital educational resources. Teachers need to tell the unique information that students will not be able to find on the Internet and transmit it in such a way that would be interesting to students (Lai, Khaddage & Knezek, 2017). The status of a teacher is directly linked to the roles they assume. As traditional roles transform in the modern environment, new roles associated with increased workload and responsibilities are emerging, which teachers must adapt to

259754022

in order to stay relevant and effective (Table 3). Lecturer Transfer information to students in the classroom or remotely using online environment, presentations, and digital educational resources 2 Scientist Research, participation in conferences of various levels 3 Ethical teacher Education of students in accordance with traditional values, norms, and attitudes 4 Moderator Creation of electronic resources, further work with them and support of students in the online environment 5 Curator Assistance and support for students in everyday life. The teacher takes the position of an assistant in the search for solutions, helps students overcome everyday difficulties, monitors their attendance, etc. 6 Methodologist Development of author's approaches and concepts of conducting classes 7 Businessman Fundraising (grants, business agreements, educational services) 8 Technical specialist Equipment maintenance during educational process, equipment adjustment, work in various computer programs 9 Consultant Counseling -mentoring outside the educational process 10 Navigator Assistance in the compilation of information on the Internet CONCLUSION The study results reveal that the majority of respondents consider mentoring as the most crucial role of a teacher in today's educational landscape. The traditional method of transferring theoretical knowledge is losing ground due to the digitalization of education. Teachers are now focusing on helping students navigate the vast amount of information available and apply their knowledge practically. Despite the perceived decline in the prestige of the teaching profession, respondents remain optimistic about the future, recognizing that the mastery of digital tools can enhance their professional activities. The transformation of teaching roles, both traditional and new, is directly linked to the digital

259754022

age and its impact on the educational process.

259754022

The bound on chaos for closed strings in Anti-de Sitter black hole backgrounds We perform a systematic study of the maximum Lyapunov exponent values λ for the motion of classical closed strings in Anti-de Sitter black hole geometries with spherical, planar and hyperbolic horizons. Analytical estimates from the linearized varia- tional equations together with numerical integrations predict the bulk Lyapunov exponent value as λ ≈ 2πTn, where n is the winding number of the string. The celebrated bound on chaos stating that λ ≤ 2πT is thus systematically modified for winding strings in the bulk. Within gauge/string duality, such strings apparently correspond to complicated operators which either do not move on Regge trajectories, or move on subleading trajectories with an unusual slope. Depending on the energy scale, the out-of-time-ordered correlation functions of these operators may still obey the bound 2πT, or they may violate it like the bulk exponent. We do not know exactly why the bound on chaos can be modified but the indication from the gauge/string dual viewpoint is that the correlation functions of the dual gauge operators never factorize and thus the original derivation of the bound on chaos does not apply. Introduction Sharp results like inequalities and no-go theorems are often the cornerstones of our understanding of physical phenomena. Besides being appealing and captivating, they are easy to test as they provide a sharp prediction on a certain quantity, and we can often learn a lot by understanding the cases when such bounds need to be generalized or abandoned. The upper bound

145856149

on the Lyapunov exponent (the rate of the growth of chaos), derived in [1] inspired by hints found in several earlier works [2][3][4][5][6][7], is an example of such a result, which is related to the dynamics of nonstationary correlation functions and provides insight into the deep and important problem of thermalization and mixing in strongly coupled systems. JHEP12(2019)150 It is clear, as discussed also in the original paper [1], that there are cases when the bound does not apply: mainly systems in which the correlation functions do not factorize even at arbitrarily long times, and also systems without a clear separation of short timescales (or collision times) and long timescales (or scrambling times). A concrete example of bound violation was found in [8] for a semiclassical system with a conserved angular momentum (inspired by the Sachdev-Ye-Kitaev (SYK) model [9][10][11][12]) and in [13], again for a SYKinspired system. In the former case, the reason is clear: the orbits that violate the bound are precisely those that cannot be treated semiclassically, so the violation just signals that the model used becomes inaccurate; in the latter case things are more complicated and the exact reason is not known. Finally, in [14] systematic higher-order quantum corrections to the bound are considered. The bound is in any case a very useful benchmark, which can tell us something on long-term dynamics of the system at hand, i.e. if some bound-violating mechanisms are at work or not. Although the bound on chaos is mainly formulated for field theories in flat spacetime, it has

145856149

an intimate connection to gravity: the prediction is that fields with gravity duals saturate the bound. This makes dynamics in asymptotically anti-de Sitter (AdS) spacetimes with a black hole particularly interesting: they have a field theory dual, 1 and black holes are conjectured to be the fastest scramblers in nature [2,3], i.e., they minimize the time for the overlap between the initial and current state to drop by an order of magnitude. Some tests of the bound for the motion of particles in the backgrounds of AdS black holes and an additional external potential were already made [15]; the authors find that the bound is systematically modified for particles hovering at the horizon and interacting with higher spin external fields. When the external field becomes scalar, the exact bound by Maldacena, Shenker and Stanford is recovered (as shown also in [16]). The idea of this paper is to study the bound on chaos in the context of motion of strings in AdS black hole geometries. Asymptotically AdS geometry is helpful not only because of the gauge/gravity duality, but also for another reason: AdS asymptotics provide a regulator, i.e., put the system in a box, making its dynamics more interesting (in asymptotically flat space, most orbits immediately escape to infinity with no opportunity to develop chaos). Now why consider strings instead of geodesics? Because geodesics are not the best way to probe the chaos generated by black holes: we know that geodesics in AdS-Schwarzschild, AdS-Reissner-Nordstrom and AdS-Kerr backgrounds (and also in all axially symmetric and static black

145856149

hole geometries) are integrable, and yet, since the horizon in all these cases has a finite Hawking temperature, there should be some thermalization and chaos going on. The logical decision is therefore to go for string dynamics, which is practically always nonintegrable in the presence of a black hole. We look mainly at the Lyapunov exponents and how they depend on the Hawking temperature. We will see that the bound of [1] is surprisingly relevant here, even though the bound was formulated for field theories with a classical gravity dual, whereas we look at the bulk dynamics of strings, which go beyond the realm of Einstein gravity. At first glance, their Lyapunov exponents should not saturate (let alone violate) the bound; in fact, at first glance, it is not JHEP12(2019)150 obvious at all how to relate the Lyapunov exponent of classical bulk orbits to the result [1], which defines the Lyapunov exponent in terms of the out-of-time ordered correlation functions (OTOC). 2 An important discovery in relation to this issue was made in [17], where the authors consider a holographically more realistic string (open string dual to a quark in Brownian motion in a heath bath), compute the Lyapunov exponent in dual field theory, and find that it exactly saturates the bound. However, their world-sheet theory, i.e., their induced metric itself looks somewhat like gravity on AdS 2 ; therefore, close connection to the Einstein gravity result is understandable. Our situation is different not only because the ring string configurations have worldsheet actions very different from

145856149

Einstein gravity but also because we look mainly at the Lyapunov exponents of the bulk orbits. 3 We will eventually look also at the OTOC in dual field theory and find that the "quantum" Lyapunov exponents do not in general coincide with the classical bulk values. However, the subject of OTOC functions is more complicated as it requires one to consider the backreaction on the background, and studying the behavior of the ring string in such backreacted geometry is in general more difficult than for the open string od [17]. Therefore, we mostly leave the OTOC and quantum Lyapunov exponent for future work. At this point we come to another question, distinct but certainly related to the chaos bound: the story of (non)integrability in various curved spacetimes. For point particles (i.e., motion on geodesics) it is usually not so difficult to check for integrability, and symmetries of the problem usually make the answer relatively easy. However, integrability in string theory remains a difficult topic. Most systematic work was done for top-down backgrounds, usually based on the differential Galois theory whose application for string integrability was pioneered in [19]. Systematic study for various top-down configurations was continued in [20][21][22]; [21] in particular provides the results for strings in a broad class of brane backgrounds, including Dp-brane, NS1 and NS5 brane configurations. The bottom line is that integrable systems are few and far apart, as could be expected. Certainly, AdS 5 × S 5 is an integrable geometry, as could be expected from its duality to the (integrable)

145856149

supersymmetric Yang-Mills field theory. In fact, direct product of AdS space and a sphere is integrable in any dimension, which is obvious from the separability of the coordinates. But already a marginal deformation destroys integrability; a specific example was found analytically and numerically in [23], for the β-deformation of super-Yang-Mills and its top-down dual. More information can be found, e.g., in the review [24]. The first study of integrability in a black hole background was [25], where the nonintegrability of string motion in asymptotically flat Schwarzschild black hole background was shown. In [26] the first study for an AdS black hole background (AdS-Schwarzschild) was performed, putting the problem also in the context of AdS/CFT correspondence. In [27] the work on top-down backgrounds was started, considering the strings 2 In addition, the scrambling concept of [2,[4][5][6][7] is more complex; it is about the equilibration of the black hole and its environment after something falls in. In other words, it necessarily includes the perturbation of the black hole itself. We do not take into account any backreaction so we cannot compute the scrambling time, only the Lyapunov exponent. 3 Another example where the bound is modified (by a factor of 2) in a theory that goes beyond Einstein gravity is [18]. JHEP12(2019)150 on the AdS×T 1,1 geometry generated in a self-consistent top-down way. For the topdown AdS-Sasaki-Einstein background the nonintegrability was proven analytically [19]. Finally, AdS-soliton and AdS-Reissner-Nordstrom were also found to be nonintegrable in [28,29]. So most well-known in AdS/CFT have nonintegrable string dynamics: AdS-Schwarzschild, AdS-Reissner-Nordstrom, AdS

145856149

soliton and AdS-Sasaki-Einstein. 4 Other results on (non)integrability can be found in [30][31][32][33]; the list is not exhaustive. Apart from the usual spherical static black holes (neutral and charged), we consider also non-spherical horizons with constant curvature. Among them are also the zero-curvature black branes, with infinite planar horizons, which are most popular in applied holography. But it is known that more general horizons can be embedded in AdS space (in general not in Minkowski space). Such black holes are usually called topological black holes, first constructed in [34][35][36][37] and generalized in [38]. The term topological is in fact partly misleading, as the backgrounds considered in some of the original papers [35] and also in our paper are not necessarily of higher topological genus: besides spherical and planar horizons, we mainly consider an infinite, topologically trivial hyperbolic horizon with constant negative curvature (pseudosphere). 5 The reader might wonder how important the non-spherical black holes are from the physical viewpoint. In fact, as shown in the aforementioned references, they arise naturally in spaces with negative cosmological constant, i.e., in AdS spaces, for example in the collapse of dust [39], and the topological versions are easily obtained through suitable gluings (identifications of points on the orbit of some discrete subgroup of the total symmetry group) of the planar or pseudospherical horizon. Another mechanism is considered in [34], where topological black holes are pair-created from instanton solutions of the cosmological C-metric (describing a pair of black holes moving with uniform acceleration). More modern work on constant-curvature black holes and

145856149

some generalizations can be found in [40][41][42], and AdS/CFT correspondence was applied to topological black holes in [43]. But our main motivation for considering non-spherical black holes is methodological, to maximally stretch the testing ground for the chaos bound and to gain insight into various chaos-generating mechanisms. In hindsight, we find that hyperbolic are roughly speaking most chaotic, because moving on a manifold of negative curvature provides an additional chaos-generating mechanism, in addition to the black hole. The plan of the paper is the following. In the next section we write down the equations of motion for a closed string in static black hole background, inspect the system analytically and numerically and show that dynamics is generically non-integrable. In the third section we compute the Lyapunov exponents numerically and estimate them analytically, formulating a generalized bound in terms of the local temperature and the string winding number. The fourth section is a rather speculative attempt to put our results in the context of the dual field theory and the derivation of the original bound from [1]; we will also try to clarify the relation of the bulk classical Lyapunov exponent to the decay rates of OTOC functions in dual field theory. The last section sums up the conclusions. 4 In [26,29] it was shown that Reissner-Nordstrom black holes in asymptotically flat space are also nonintegrable. 5 In fact, constant-curvature black holes would be a more suitable term than topological black holes. where dσ 2 N −1 is the horizon manifold, which has curvature k, and m

145856149

and q define the mass and charge of the black hole. It is a vacuum solution of the Einstein equations with constant negative cosmological constant and thus interpolates to AdS space with radius 1. From now on let us stick to N = 3 unless specified otherwise. For k = 1 we have the familiar spherical black hole. For k = 0 we get the planar horizon (black brane) popular in AdS/CFT applications. 6 Finally, for k = −1 the horizon is an infinite hyperbolic sheet (pseudosphere), with the symmetry group SO(2, 1). 7 Notice that k can always be rescaled together with the coordinates on σ 2 thus we only consider k = −1, 0, 1. The metric of the horizon surface takes the form with sink(x) = sin x for k = 1, sink(x) = x for k = 0 and sink(x) = sinh(x) for k = −1. A closed string with tension 1/α on the worldsheet (τ, σ) with target space X µ and the metric G µν is described by the Polyakov action: In our black hole backgrounds we always have B µν = 0 so we can pick the gauge h ab = η ab = diag(−1, 1). This gives the Virasoro constraints where we introduce the notationẊ ≡ ∂ τ X, X ≡ ∂ σ X. The first constraint is the Hamiltonian constraint H = 0. We consider closed strings, so 0 ≤ σ ≤ 2π. From the second constraint the following ansatz is consistent (of course, it is not

145856149

the only one possible): We denote the (dynamical) target-space coordinates X µ (τ, σ) by capital letters T , R, Φ 1 , Φ 2 , to differentiate them from the notation for spacetime coordinates t, r, φ 1 , φ 2 in the metric (2.1). The form (2.5) was tried in most papers exploring the integrability and chaos JHEP12(2019)150 of strings [19,[25][26][27][28][29]. It is not an arbitrary ansatz: the winding of Φ 2 follows from the equations of motion, i.e., from the fact that Φ 2 is a cyclic coordinate, leading to the solution Φ 2 = 0. Since Φ 2 has trivial dynamics, from now on we will denote Φ ≡ Φ 1 . The equations of motion follow from (2.3): Clearly, the stationarity of the metric yields the first integral E with the informal meaning of mechanical energy for the motion along the R and Φ coordinates (it is not the total energy in the strict sense). The system is more transparent in Hamiltonian form, with the canonical momenta the second equality being the Virasoro constraint. We thus have a 2-degrees-of-freedom system (due to the integral of motion E, i.e., the cyclic coordinate T ), with a constraint, effectively giving a 1.5-degrees-of-freedom system, moving on a three-dimensional manifold in the phase space (R, P R , Φ, P Φ ). Notice that the motion along a geodesic is obtained for n = 0; in this case, the system is trivially separable and becomes just motion in a central potential. For nonzero n, the

145856149