Protein docking using surface matching and supervised machine learning

Computational prediction of protein complex structures through docking offers a means to gain a mechanistic understanding of protein interactions that mediate biological processes. This is particularly important as the number of experimentally determined structures of isolated proteins exceeds the number of structures of complexes. A comprehensive docking procedure is described in which efficient sampling of conformations is achieved by matching surface normal vectors, fast filtering for shape complementarity, clustering by RMSD, and scoring the docked conformations using a supervised machine learning approach. Contacting residue pair frequencies, residue propensities, evolutionary conservation, and shape complementarity score for each docking conformation are used as input data to a Random Forest classifier. The performance of the Random Forest approach for selecting correctly docked conformations was assessed by cross-validation using a nonredundant benchmark set of X-ray structures for 93 heterodimer and 733 homodimer complexes. The single highest rank docking solution was the correct (near-native) structure for slightly more than one third of the complexes. Furthermore, the fraction of highly ranked correct structures was significantly higher than the overall fraction of correct structures, for almost all complexes. A detailed analysis of the difficult to predict complexes revealed that the majority of the homodimer cases were explained by incorrect oligomeric state annotation. Evolutionary conservation and shape complementarity score as well as both underrepresented and overrepresented residue types and residue pairs were found to make the largest contributions to the overall prediction accuracy. Finally, the method was also applied to docking unbound subunit structures from a previously published benchmark set.     

Strain design optimization using reinforcement learning

Engineered microbial cells present a sustainable alternative to fossil-based synthesis of chemicals and fuels. Cellular synthesis routes are readily assembled and introduced into microbial strains using state-of-the-art synthetic biology tools. However, the optimization of the strains required to reach industrially feasible production levels is far less efficient. It typically relies on trial-and-error leading into high uncertainty in total duration and cost. New techniques that can cope with the complexity and limited mechanistic knowledge of the cellular regulation are called for guiding the strain optimization.In this paper, we put forward a multi-agent reinforcement learning (MARL) approach that learns from experiments to tune the metabolic enzyme levels so that the production is improved. Our method is model-free and does not assume prior knowledge of the microbe’s metabolic network or its regulation. The multi-agent approach is well-suited to make use of parallel experiments such as multi-well plates commonly used for screening microbial strains.We demonstrate the method’s capabilities using the genome-scale kinetic model of Escherichia coli, k-ecoli457, as a surrogate for an in vivo cell behaviour in cultivation experiments. We investigate the method’s performance relevant for practical applicability in strain engineering i.e. the speed of convergence towards the optimum response, noise tolerance, and the statistical stability of the solutions found. We further evaluate the proposed MARL approach in improving L-tryptophan production by yeast Saccharomyces cerevisiae, using publicly available experimental data on the performance of a combinatorial strain library.Overall, our results show that multi-agent reinforcement learning is a promising approach for guiding the strain optimization beyond mechanistic knowledge, with the goal of faster and more reliably obtaining industrially attractive production levels.     

Parameter optimization model of learning in stepping motion

In this study we combine the representation of motion by a finite number of hardwired functions with parameter optimization to model learning during a stepping motion. Representation of experimental kinematic data by a finite number of predetermined functions and undetermined coefficients was analyzed. Least squares approximation was used to represent experimental data of stepping motions over obstacles of different heights. Functional relationships between coefficients and obstacles heights were also obtained. Learning of stepping over an obstacle was then formulated as a finite dimensional optimization problem. The pattern of foot path, and joint angles trajectories obtained by this learning model, were then compared to the experimental data. The results of the data fitting analysis and of the optimization process as a model for motion learning, indicate that motion can be adequately represented by a set of hardwired functions, and a finite number of task dependent coefficients.     

Distributed scalable multi-robot learning using particle swarm optimization

Designing effective behavioral controllers for mobile robots can be difficult and tedious; this process can be circumvented by using online learning techniques which allow robots to generate their own controllers online in an automated fashion. In multi-robot systems, robots operating in parallel can potentially learn at a much faster rate by sharing information amongst themselves. In this work, we use an adapted version of the Particle Swarm Optimization algorithm in order to accomplish distributed online robotic learning in groups of robots with access to only local information. The effectiveness of the learning technique on a benchmark task (generating high-performance obstacle avoidance behavior) is evaluated for robot groups of various sizes, with the maximum group size allowing each robot to individually contain and manage a single PSO particle. To increase the realism of the technique, different PSO neighborhoods based on limitations of real robotic communication are tested and compared in this scenario. We explore the effect of varying communication power for one of these communication-based PSO neighborhoods. To validate the effectiveness of these learning techniques, fully distributed online learning experiments are run using a group of 10 real robots, generating results which support the findings from our simulations.     

Sensitive period for sensorimotor integration during vocal motor learning

Sensory experience during sensitive periods in development may direct the organization of neural substrates, thereby permanently influencing subsequent adult behavior. We report a sensitive period during the imitative motor learning phase of sensorimotor integration in birdsong development. By temporarily and reversibly blocking efference to the vocal muscles, we disrupted vocal motor practice during selected stages of song development. Motor disruption during prolonged periods early in development, which allows recovery of vocal control prior to the onset of adult song, has no effect on adult song production. However, song disruption late in development, during the emergence of adult song, results in permanent motor defects in adult song production. These results reveal a decreased ability to compensate for interference with motor function when disturbances occur during the terminal stage of vocal motor development. Temporary disruption of syringeal motor control in adults does not produce permanent changes in song production. Permanent vocal aberrations in juveniles are evident exclusively in learned song elements rather than nonlearned calls, suggesting that the sensitive period is associated with motor learning.     

Physical simulation of laryngeal disorders using a multiple-mass vocal fold model

This paper reports on the use of a high-dimensional discrete vocal fold model for the simulation of voice production under the presence of laryngeal disorders. Specifically, the effect of increases in mass and stiffness, both unilateral and bilateral, has been analysed independently for both magnitudes. The glottal flow waveform and the mass displacement have been studied and the obtained results are coherent with clinical observations that relate mass increments with lowered fundamental frequencies and mass and stiffness increments with reduced vibratory amplitudes of vocal folds. The reported results also indicate that asymmetries in the physical properties of vocal folds result in asymmetries in their vibratory patterns, including phase, amplitude and behaviour on collision. These are also correlated with voice perturbation measures such as jitter, shimmer and normalised noise energy.     

Application-oriented deep learning model for early warning of rice blast in Taiwan

Rice blast is one of the most devastating diseases that threatens rice production in Taiwan. A rice blast forecasting model is required to guide the precise application of fungicide. Therefore, BlastGRU-TW model based on deep learning algorithms was developed in this study. The input data comprised approximately 1000 rice blast surveys, collected from 50 fields throughout Taiwan between 2014 and 2021, and the corresponding weather data retrieved from weather observation network in Taiwan. Common and easily accessible meteorological factors, i.e. temperature, humidity, precipitation, and wind, were converted into 20 daily meteorological features which were coupled with different time intervals between 1 and 30 days before survey (DBS) to train the model. The results showed that seven meteorological features (daily maximum temperature, daily mean temperature, daily minimum temperature, daily mean humidity, daily mean WV and daily mean Wu) and the interval from −4 to −24 DBS were informative in disease prediction, thereby indicating that the proposed model could predict the risk of rice blast by using meteorological data 4–24 days before new disease symptoms appeared. The proposed BlastGRU-TW model achieved an accuracy of 87.3%. Furthermore, on adding the 3 day forecast weather data from Weather Research and Forecasting (WRF) model in the proposed model, the forecast extended to 7 days ahead of the appearance of new symptoms. Moreover, the BLASTAM model developed in Japan was implemented and validated in Taiwan to evaluate its applicability in different geographical areas. Finally, a rice blast early warning system (https://mycolab.pp.nchu.edu.tw/blast_forecast/index_en.php) equipped with an interactive web-based map is now available for real-time forecasting of the risk of rice blast in paddy field across Taiwan.     

Auditory feedback in learning and maintenance of vocal behaviour

Songbirds are one of the best-studied examples of vocal learners. Learning of both human speech and birdsong depends on hearing. Once learned, adult song in many species remains unchanging, suggesting a reduced influence of sensory experience. Recent studies have revealed, however, that adult song is not always stable, extending our understanding of the mechanisms involved in song maintenance, and their similarity to those active during song learning. Here we review some of the processes that contribute to song learning and production, with an emphasis on the role of auditory feedback. We then consider some of the possible neural substrates involved in these processes, particularly basal ganglia circuitry. Although a thorough treatment of human speech is beyond the scope of this article, we point out similarities between speech and song learning, and ways in which studies of these disparate behaviours complement each other in developing an understanding of general principles that contribute to learning and maintenance of vocal behaviour.     

Use of multistate Bennett acceptance ratio method for free-energy calculations from enhanced sampling and free-energy perturbation

Multistate Bennett acceptance ratio (MBAR) works as a method to analyze molecular dynamics (MD) simulation data after the simulations have been finished. It is widely used to estimate free-energy changes between different states and averaged properties at the states of interest. MBAR allows us to treat a wide range of states from those at different temperature/pressure to those with different model parameters. Due to the broad applicability, the MBAR equations are rather difficult to apply for free-energy calculations using different types of MD simulations including enhanced conformational sampling methods and free-energy perturbation. In this review, we first summarize the basic theory of the MBAR equations and categorize the representative usages into the following four: (i) perturbation, (ii) scaling, (iii) accumulation, and (iv) full potential energy. For each, we explain how to prepare input data using MD simulation trajectories for solving the MBAR equations. MBAR is also useful to estimate reliable free-energy differences using MD trajectories based on a semi-empirical quantum mechanics/molecular mechanics (QM/MM) model and ab initio QM/MM energy calculations on the MD snapshots. We also explain how to use the MBAR software in the GENESIS package, which we call mbar_analysis, for the four representative cases. The proposed estimations of free-energy changes and thermodynamic averages are effective and useful for various biomolecular systems.     

Combination treatment optimization using a pan-cancer pathway model

The design of efficient combination therapies is a difficult key challenge in the treatment of complex diseases such as cancers. The large heterogeneity of cancers and the large number of available drugs renders exhaustive in vivo or even in vitro investigation of possible treatments impractical. In recent years, sophisticated mechanistic, ordinary differential equation-based pathways models that can predict treatment responses at a molecular level have been developed. However, surprisingly little effort has been put into leveraging these models to find novel therapies. In this paper we use for the first time, to our knowledge, a large-scale state-of-the-art pan-cancer signaling pathway model to identify candidates for novel combination therapies to treat individual cancer cell lines from various tissues (e.g., minimizing proliferation while keeping dosage low to avoid adverse side effects) and populations of heterogeneous cancer cell lines (e.g., minimizing the maximum or average proliferation across the cell lines while keeping dosage low). We also show how our method can be used to optimize the drug combinations used in sequential treatment plans—that is, optimized sequences of potentially different drug combinations—providing additional benefits. In order to solve the treatment optimization problems, we combine the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) algorithm with a significantly more scalable sampling scheme for truncated Gaussian distributions, based on a Hamiltonian Monte-Carlo method. These optimization techniques are independent of the signaling pathway model, and can thus be adapted to find treatment candidates for other complex diseases than cancers as well, as long as a suitable predictive model is available.     

利用文献精读教学新模式优化遗传学教学

《遗传学》是生命科学相关专业本科阶段最重要的课程之一。近年来,随着生命科学领域研究的不断深入,新知识与新技术也在不断更新。但遗传学的教学模式目前仍以理论讲授为主,这使得抽象的原理难以被学生理解接受,直接影响了教学效果。因此探索新的教学模式尤为必要。2010年以来我校在生物技术专业《微生物遗传学》教学中开展了新教学模式——文献精读,文章从文献精读的前期课程基础,如何选择专业文献,怎样组织教学过程,开展文献精读对学生和教师的意义等方面全面分析了实施情况和应用价值,指出该教学模式体现了“前沿”和“经典”的结合,使书本的知识在实践中具体化,既提高学生的学习效果,激发学习兴趣,又开拓了学生的思路,锻炼其能力。这种教学模式为《遗传学》教学授课不断探索新的模式、在“精准医疗”时代下如何培养兼具临床与科研能力的医疗人才提供新思路。     

The role of auditory feedback in vocal learning and maintenance

Auditory experience is critical for the acquisition and maintenance of learned vocalizations in both humans and songbirds. Despite the central role of auditory feedback in vocal learning and maintenance, where and how auditory feedback affects neural circuits important to vocal control remain poorly understood. Recent studies of singing birds have uncovered neural mechanisms by which feedback perturbations affect vocal plasticity and also have identified feedback-sensitive neurons at or near sites of auditory and vocal motor interaction. Additionally, recent studies in marmosets have underscored that even in the absence of vocal learning, vocalization remains flexible in the face of changing acoustical environments, pointing to rapid interactions between auditory and vocal motor systems. Finally, recent studies show that a juvenile songbird's initial auditory experience of a song model has long-lasting effects on sensorimotor neurons important to vocalization, shedding light on how auditory memories and feedback interact to guide vocal learning.     

A blueprint for vocal learning: auditory predispositions from brains to genomes

Memorizing and producing complex strings of sound are requirements for spoken human language. We share these behaviours with likely more than 4000 species of songbirds, making birds our primary model for studying the cognitive basis of vocal learning and, more generally, an important model for how memories are encoded in the brain. In songbirds, as in humans, the sounds that a juvenile learns later in life depend on auditory memories formed early in development. Experiments on a wide variety of songbird species suggest that the formation and lability of these auditory memories, in turn, depend on auditory predispositions that stimulate learning when a juvenile hears relevant, species-typical sounds. We review evidence that variation in key features of these auditory predispositions are determined by variation in genes underlying the development of the auditory system. We argue that increased investigation of the neuronal basis of auditory predispositions expressed early in life in combination with modern comparative genomic approaches may provide insights into the evolution of vocal learning.     

Dialect change in resident killer whales: implications for vocal learning and cultural transmission

Variation in vocal signals among populations and social groups of animals provides opportunities for the study of the mechanisms of behavioural change and their importance in generating and maintaining behavioural variation. We analysed two call types made by two matrilineal social groups of resident killer whales, Orcinus orca, over 12-13 years. We used a neural network-based index of acoustic similarity to identify mechanisms of call differentiation. A test for structural modification of the calls detected significant changes in one call type in both groups, but not in the other. For the modified call type, the rate of divergence between the two groups was significantly lower than the rate of modification within either group showing that calls were modified in a similar fashion in the two groups. An analysis of structural parameters detected no strong directionality in the change. The pattern of call modification could have been caused by maturational changes to the calls or, if killer whale dialects are learned behavioural traits, cultural drift in the structure of the calls together with horizontal transmission of modifications between the two groups. Such vocal matching between members of different matrilines would suggest that vocal learning is not limited to vertical transmission from mother to offspring, which has important implications for models of gene-culture coevolution. Copyright 2000 The Association for the Study of Animal Behaviour.     

Song divergence and male dispersal among bird populations: a spatially explicit model testing the role of vocal learning

The importance of bird song in sexual selection and the widespread existence of geographical variation in song has raised the question of whether dialects can promote the evolution of new, reproductively isolated, subgroups. However, the effect of song learning on the relation between song divergence and gene flow is still poorly understood. We explored the origin and maintenance of song divergence and its consequences for male dispersal between populations. To do this, we used a spatially explicit model to study the effect of vocal learning mode (predispersal and postdispersal) and the role of intrasexual selection among males. The majority of scenarios simulated in the model led to significant genetic and phenotypic song divergence, consistent with the widespread occurrence of dialects in natural populations. Intrasexual selection in the form of male exclusion, dependent on song matching between neighbours, was an efficient mechanism driving song divergence. Most importantly, song learning was decisive in causing an incongruence of genetic divergence and song divergence between populations. As a consequence, song learning often prevented the expected negative relation between the degree of song divergence and dispersal. These findings may help to interpret empirical data on song divergence and gene flow and provide qualitative and testable predictions for conditions under which intraspecific song variation may promote reproductive divergence and speciation. Copyright 2003 Published by Elsevier Science Ltd on behalf of The Association for the Study of Animal Behaviour.      

Song evolution,speciation, and vocal learning in passerine birds

Phenotypic divergence can promote reproductive isolation and speciation, suggesting a possible link between rates of phenotypic evolution and the tempo of speciation at multiple evolutionary scales. To date, most macroevolutionary studies of diversification have focused on morphological traits, whereas behavioral traits─including vocal signals─are rarely considered. Thus, although behavioral traits often mediate mate choice and gene flow, we have a limited understanding of how behavioral evolution contributes to diversification. Furthermore, the developmental mode by which behavioral traits are acquired may affect rates of behavioral evolution, although this hypothesis is seldom tested in a phylogenetic framework. Here, we examine evidence for rate shifts in vocal evolution and speciation across two major radiations of codistributed passerines: one oscine clade with learned songs (Thraupidae) and one suboscine clade with innate songs (Furnariidae). We find that evolutionary bursts in rates of speciation and song evolution are coincident in both thraupids and furnariids. Further, overall rates of vocal evolution are higher among taxa with learned rather than innate songs. Taken together, these findings suggest an association between macroevolutionary bursts in speciation and vocal evolution, and that the tempo of behavioral evolution can be influenced by variation in developmental modes among lineages.     

Analysis of the force-sharing problem using an optimization model

In biomechanics, one frequently used approach for finding a unique set of muscle forces in the 'force-sharing problem' is to formulate and solve a non-linear optimization problem of the form: min phi(f)= summation operator (f(i)/omega(i))(alpha) subject to Af = b and f > or = 0. Solutions to this problem have typically been obtained numerically for complex models, or analytically for specific musculoskeletal geometries. Here, we present simple geometrical methods for analyzing the solution to this family of optimization problems for a general n-degrees-of-freedom musculoskeletal system. For example, it is shown that the moment-arm vectors of active (f(i) > 0) and passive (f(i) = 0) muscles are separated by a hyperplane through the origin of the moment-arm vector space. For the special case of a system with two degrees-of-freedom, solutions can be readily represented in graphical form. This allows for powerful interpretations of force-sharing calculated using optimization.