MSMBuilder: Statistical Models for Biomolecular Dynamics

MSMBuilder is a software package for building statistical models of high-dimensional time-series data. It is designed with a particular focus on the analysis of atomistic simulations of biomolecular dynamics such as protein folding and conformational change. MSMBuilder is named for its ability to construct Markov state models (MSMs), a class of models that has gained favor among computational biophysicists. In addition to both well-established and newer MSM methods, the package includes complementary algorithms for understanding time-series data such as hidden Markov models and time-structure based independent component analysis. MSMBuilder boasts an easy to use command-line interface, as well as clear and consistent abstractions through its Python application programming interface. MSMBuilder was developed with careful consideration for compatibility with the broader machine learning community by following the design of scikit-learn. The package is used primarily by practitioners of molecular dynamics, but is just as applicable to other computational or experimental time-series measurements.     

A two-step procedure for coupling development and usage of a pair of human neck models

Both finite element models and multi-body models of human head-neck complex had been widely used in neck injuries analysis, as the former could be used to generate detailed stress strain information and the later could generate dynamic responses with high efficiency. Sometimes, detailed stress and strain information were hoped to be obtained more efficiently, but current methods were not effective enough when they were used to analyze responses of human head neck complex to long duration undulate accelerations. In this paper, a two-step procedure for ‘parallel’ development and ‘sequential’ usage of a pair of human head neck models was discussed. The pair of models contained a finite element model and a multi-body model, which were developed based on the coupling ‘parallel’ procedure using the same bio-realistic geometry. After being validated using available data, the pair of human neck models were applied to analyze biomechanical responses of pilot’s neck during arrested landing operation according to the ‘sequential’ procedure, because typical sustained undulate accelerations usually appeared during such processes. The results, including both kinematic and detailed biomechanical responses of human head-neck complex, were obtained with preferable efficiency. This research provided an effective way for biomechanical analysis of human head neck responses to sustained undulate accelerations.     

A computer-based simulation of childbirth using the partial Dirichlet–Neumann contact method with total Lagrangian explicit dynamics on the GPU

During physiological or ‘natural’ childbirth, the fetal head follows a distinct motion pattern—often referred to as the cardinal movements or ‘mechanisms’ of childbirth—due to the biomechanical interaction between the fetus and maternal pelvic anatomy. The research presented in this paper introduces a virtual reality-based simulation of physiological childbirth. The underpinning science is based on two numerical algorithms including the total Lagrangian explicit dynamics method to calculate soft tissue deformation and the partial Dirichlet–Neumann contact method to calculate the mechanical contact interaction between the fetal head and maternal pelvic anatomy. The paper describes the underlying mathematics and algorithms of the solution and their combination into a computer-based implementation. The experimental section covers first a number of validation experiments on simple contact mechanical problems which is followed by the main experiment of running a virtual reality childbirth. Realistic mesh models of the fetus, bony pelvis and pelvic floor muscles were subjected to the intra-uterine expulsion forces which aim to propel the virtual fetus through the virtual birth canal. Following a series of simulations, taking variations in the shape and size of the geometric models into account, we consistently observed the cardinal movements in the simulator just as they happen in physiological childbirth. The results confirm the potential of the simulator as a predictive tool for problematic childbirths subject to patient-specific adaptations.

     

‘Ghost’ experiments and the dissection of social learning in humans and animals

The focus of this review is the experimental techniques used to identify forms of social learning shown by humans and nonhuman animals. Specifically, the ‘ghost display’ and ‘end‐state’ conditions, which have been used to tease apart imitative and emulative learning are evaluated. In a ghost display, the movements of an apparatus are demonstrated, often through the discrete use of fishing‐line or hidden mechanisms, without a live model acting directly upon the apparatus so that the apparatus appears to be operated as if by a ‘ghostly’ agent. In an end‐state condition, an observing individual is shown the initial state of the test apparatus, the apparatus is then manipulated out‐of‐sight and then represented to the individual in its final state. The aim of the ghost display condition is to determine whether individuals are able to emulate by replicating the movements of an apparatus, or perform a task, without requiring information about the bodily movements required to do so (imitation). The end‐state condition is used to identify goal‐emulation by assessing whether the observer can replicate the steps required to solve the task without having been shown the required body actions or task movements. The responses of individuals tested with either the ghost display and/or end‐state conditions are compared to those of further individuals who have observed a full demonstration by either a human experimenter or a conspecific. The responses of a control group, to whom no information has been provided about the test apparatus or required actions, are also compared and evaluated. The efficacy of these experimental techniques employed with humans, nonhuman primates, dogs, rats and birds are discussed and evaluated. The experiments reviewed herein emphasise the need to provide ghost displays and end‐state conditions in combination, along with full live demonstrations and a no‐information control. Future research directions are proposed.     

Collection of antimicrobial peptides database and its derivatives: Applications and beyond

Collection of antimicrobial peptides (CAMP), CAMPSign, and ClassAMP are open‐access resources that have been developed to enhance research on antimicrobial peptides (AMPs). Comprehensive information on AMPs and machine learning‐based predictive models are made available for users through these resources. As of date, CAMP_R3 has 10,247 sequences, 757 structures, and 114 family‐specific signatures of AMPs along with associated tools for AMP sequence and structure analysis. CAMPSign uses family‐specific sequence conservation, in the form of patterns and hidden Markov models for identification of AMPs. ClassAMP can be used to classify AMPs as antibacterial, antifungal, or antiviral based on sequence information. Here we describe CAMP and its derivatives and illustrate, with a few examples, the contribution of these online resources to the advancement of our current understanding of AMPs.     

Parametric analysis of the biomechanical response of head subjected to the primary blast loading – a data mining approach

Traumatic brain injury due to primary blast loading has become a signature injury in recent military conflicts and terrorist activities. Extensive experimental and computational investigations have been conducted to study the interrelationships between intracranial pressure response and intrinsic or ‘input’ parameters such as the head geometry and loading conditions. However, these relationships are very complicated and are usually implicit and ‘hidden’ in a large amount of simulation/test data. In this study, a data mining method is proposed to explore such underlying information from the numerical simulation results. The heads of different species are described as a highly simplified two-part (skull and brain) finite element model with varying geometric parameters. The parameters considered include peak incident pressure, skull thickness, brain radius and snout length. Their interrelationship and coupling effect are discovered by developing a decision tree based on the large simulation data-set. The results show that the proposed data-driven method is superior to the conventional linear regression method and is comparable to the nonlinear regression method. Considering its capability of exploring implicit information and the relatively simple relationships between response and input variables, the data mining method is considered to be a good tool for an in-depth understanding of the mechanisms of blast-induced brain injury. As a general method, this approach can also be applied to other nonlinear complex biomechanical systems.     

Detecting distant homologs using phylogenetic tree-based HMMs

It is often desired to identify further homologs of a family of biological sequences from the ever-growing sequence databases. Profile hidden Markov models excel at capturing the common statistical features of a group of biological sequences. With these common features, we can search the biological database and find new homologous sequences. Most general profile hidden Markov model methods, however, treat the evolutionary relationships between the sequences in a homologous group in an ad-hoc manner. We hereby introduce a method to incorporate phylogenetic information directly into hidden Markov models, and demonstrate that the resulting model performs better than most of the current multiple sequence-based methods for finding distant homologs.     

ToPS: A Framework to Manipulate Probabilistic Models of Sequence Data

Discrete Markovian models can be used to characterize patterns in sequences of values and have many applications in biological sequence analysis, including gene prediction, CpG island detection, alignment, and protein profiling. We present ToPS, a computational framework that can be used to implement different applications in bioinformatics analysis by combining eight kinds of models: (i) independent and identically distributed process; (ii) variable-length Markov chain; (iii) inhomogeneous Markov chain; (iv) hidden Markov model; (v) profile hidden Markov model; (vi) pair hidden Markov model; (vii) generalized hidden Markov model; and (viii) similarity based sequence weighting. The framework includes functionality for training, simulation and decoding of the models. Additionally, it provides two methods to help parameter setting: Akaike and Bayesian information criteria (AIC and BIC). The models can be used stand-alone, combined in Bayesian classifiers, or included in more complex, multi-model, probabilistic architectures using GHMMs. In particular the framework provides a novel, flexible, implementation of decoding in GHMMs that detects when the architecture can be traversed efficiently.

This is a PLOS Computational Biology Software Article.

     

THE MUSCULAR CONTROL OF VERTEBRATE SWIMMING MOVEMENTS

• 1 The succession of hypotheses on the role of myotomal muscle in the generation of swimming movements is described and the conventional concept of ‘waves of contraction’ is shown to be based on a number of misinterpretations.
• 2 The form of undulatory movements in vertebrate swimmers is characterized by the properties of the sinusoidal oscillation of parts of the body about the axis of progression. An important variable is the relative amplitude of the lateral oscillation of the head end, which can be large in some animals though usually small in most adult aquatic vertebrates.
• 3 Cinematographic records of swimming animals are examined to determine the forces involved in the generation of waves of bending. A simplified analysis suggests that undulation can be produced by alternation of tension development from side to side without ‘waves of contraction’ passing down the body.
• 4 Model systems which are able to flex from side to side are considered and two types distinguished - the ‘resistance-dominated’ which propagates waves of bending from centre to extremities, and the ‘stiffness-dominated’ which does not. The type to which a model belongs is determined by the interrelationship of its stiffness and resistance, and the power with which it flexes.
• 5 A model homogeneous in its properties along its length cannot generate longitudinal movement by flexing from side to side. Some degree of unevenness from one end to the other is required for propulsion.
• 6 Observations of the movements of an ‘ostraciiform model’ are shown to discount previous theories of the hydromechanics of swimming by the oscillation of a stiff tail about a single pivot. A new interpretation is provided.
• 7 The majority of vertebrate swimmers behave like ‘hybrid oscillators’ which flex from side to side, ‘resistance-dominated’ posteriorly and ‘stiffness-dominated’ anteriorly.
• 8 The origin of the ‘waves of contraction’ suggested by electromyograms of swimming animals is traced to the requirement for a tail of variable stiffness for variable frequency of oscillation and to the need to reduce lateral oscillation of the head. Delayed contraction posteriorly and early contraction anteriorly contribute to these functions.
• 9 The ability of amphioxus to swim backwards and the inability of most vertebrates to do so is related to their structural organization in the form of ‘hybrid oscillators’.
• 10 Electromyograms are examined in the light of these mechanical models. A developmental sequence is described for the newt which illustrates the organization of the muscular control of swimming movements and may throw light upon the development of the neural mechanism.

     

Active inference and agency: optimal control without cost functions

This paper describes a variational free-energy formulation of (partially observable) Markov decision problems in decision making under uncertainty. We show that optimal control can be cast as active inference. In active inference, both action and posterior beliefs about hidden states minimise a free energy bound on the negative log-likelihood of observed states, under a generative model. In this setting, reward or cost functions are absorbed into prior beliefs about state transitions and terminal states. Effectively, this converts optimal control into a pure inference problem, enabling the application of standard Bayesian filtering techniques. We then consider optimal trajectories that rest on posterior beliefs about hidden states in the future. Crucially, this entails modelling control as a hidden state that endows the generative model with a representation of agency. This leads to a distinction between models with and without inference on hidden control states; namely, agency-free and agency-based models, respectively.     

Why and how to objectively evaluate medical gestures?

Junior physicians mainly learn during their observation in the operative room. The senior physicians evaluate them based on the same kind of observation. Knowledge transfer evaluation is thus done without quantitative methods and it mainly lies on a subjective assessment. In this paper, we present some recent techniques used to objectively evaluate medical gestures. The classical techniques used are Hidden Markov Models (HMM) or Dynamic Time Warping (DTW). Both techniques lies on the temporal analysis of the gestures. We proposed here a technique based on the arc length parametrization in order to analyze the gestures in space which is more appropriate because it gives information about the shape of the gestures independently of the chosen coordinate system.     

Neural networks for protein classification

This paper describes a biomolecular classification methodology based on multilayer perceptron neural networks. The system developed is used to classify enzymes found in the Protein Data Bank. The primary goal of classification, here, is to infer the function of an (unknown) enzyme by analysing its structural similarity to a given family of enzymes. A new codification scheme was devised to convert the primary structure of enzymes into a real-valued vector. The system was tested with a different number of neural networks, training set sizes and training epochs. For all experiments, the proposed system achieved a higher accuracy rate when compared with profile hidden Markov models. Results demonstrated the robustness of this approach and the possibility of implementing fast and efficient biomolecular classification using neural networks.     

Combining phylogenetic and hidden Markov models in biosequence analysis.

A few models have appeared in recent years that consider not only the way substitutions occur through evolutionary history at each site of a genome, but also the way the process changes from one site to the next. These models combine phylogenetic models of molecular evolution, which apply to individual sites, and hidden Markov models, which allow for changes from site to site. Besides improving the realism of ordinary phylogenetic models, they are potentially very powerful tools for inference and prediction--for example, for gene finding or prediction of secondary structure. In this paper, we review progress on combined phylogenetic and hidden Markov models and present some extensions to previous work. Our main result is a simple and efficient method for accommodating higher-order states in the HMM, which allows for context-dependent models of substitution--that is, models that consider the effects of neighboring bases on the pattern of substitution. We present experimental results indicating that higher-order states, autocorrelated rates, and multiple functional categories all lead to significant improvements in the fit of a combined phylogenetic and hidden Markov model, with the effect of higher-order states being particularly pronounced.     

PredSL： A Tool for the N-terminal Sequence-based Prediction of Protein Subcellular Localization

The ability to predict the subcellular localization of a protein from its sequence is of great importance, as it provides information about the protein＇s function. We present a computational tool, PredSL, which utilizes neural networks, Markov chains, profile hidden Markov models, and scoring matrices for the prediction of the subcellular localization of proteins in eukaryotic cells from the N-terminal amino acid sequence. It aims to classify proteins into five groups： chloroplast, thylakoid, mitochondrion, secretory pathway, and ＂other＂. When tested in a fivefold cross-validation procedure, PredSL demonstrates 86.7% and 87.1% overall accuracy for the plant and non-plant datasets, respectively. Compared with TargetP, which is the most widely used method to date, and LumenP, the results of PredSL are comparable in most cases. When tested on the experimentally verified proteins of the Saccharomyces cerevisiae genome, PredSL performs comparably if not better than any available algorithm for the same task. Furthermore, PredSL is the only method capable for the prediction of these subcellular localizations that is available as a stand-alone application through the URL： http：//bioinformatics.biol.uoa.gr/PredSL/.     

Rhesus monkeys correctly read the goal-relevant gestures of a human agent

When humans point, they reveal to others their underlying intent to communicate about some distant goal. A controversy has recently emerged based on a broad set of comparative and phylogenetically relevant data. In particular, whereas chimpanzees (Pan troglodytes) have difficulty in using human-generated communicative gestures and actions such as pointing and placing symbolic markers to find hidden rewards, domesticated dogs (Canis familiaris) and silver foxes (Urocyon cinereoargenteus) readily use such gestures and markers. These comparative data have led to the hypothesis that the capacity to infer communicative intent in dogs and foxes has evolved as a result of human domestication. Though this hypothesis has met with challenges, due in part to studies of non-domesticated, non-primate animals, there remains the fundamental question of why our closest living relatives, the chimpanzees, together with other non-human primates, generally fail to make inferences about a target goal of an agent's communicative intent. Here, we add an important wrinkle to this phylogenetic pattern by showing that free-ranging rhesus monkeys (Macaca mulatta) draw correct inferences about the goals of a human agent, using a suite of communicative gestures to locate previously concealed food. Though domestication and human enculturation may play a significant role in tuning up the capacity to infer intentions from communicative gestures, these factors are not necessary.     

Quasi-consensus-based comparison of profile hidden Markov models for protein sequences

A simple approach for the sensitive detection of distant relationships among protein families and for sequence-structure alignment via comparison of hidden Markov models based on their quasi-consensus sequences is presented. Using a previously published benchmark dataset, the approach is demonstrated to give better homology detection and yield alignments with improved accuracy in comparison to an existing state-of-the-art dynamic programming profile-profile comparison method. This method also runs significantly faster and is therefore suitable for a server covering the rapidly increasing structure database. A server based on this method is available at http://liao.cis.udel.edu/website/servers/modmod     

Influence of adaptation to horizontal body position on pointing errors and visual estimation of a target position in humans

The central program of a targeted movement includes a component intended for to compensate for the weight of the arm; this is why the accuracy of pointing to a memorized position of the visual target in darkness depends on orientation of the moving limb in relation to the vertical axis. Transition from the vertical to the horizontal body position is accompanied by a shift of the final hand position along the body axis towards the head. We studied how pointing errors and visual localization of the target are modified due to adaptation to the horizontal body position; targeted movements to a real target were repeatedly performed during the adaptation period. Three types of experiments were performed: a basic experiment, and two different experiments with adaptation realized under somewhat dissimilar conditions. In the course of the first adaptation experiment, subjects received no visual information on the hand’s position in space, and targeted movements of the arm to a luminous target could be corrected using proprioceptive information only. With such a paradigm, the accuracy of pointing to memorized visual targets showed no adaptation-related changes. In the second adaptation experiment, subjects were allowed to continuously view a marker (a light-emitting diode taped to the fingertip). After such adaptation practice, the accuracy of pointing movements to memorized targets increased: both constant and variational errors, as well as both components of constant error (i.e.,X andY errors) significantly dropped. Testing the accuracy of visual localization of the targets by visual/verbal adjustment, performed after this adaptation experiment, showed that the pattern of errors did not change compared with that in the basic experiment. Therefore, we can conclude that sensorimotor adaptation to the horizontal position develops much more successfully when the subject obtains visual information about the working point position; such adaptation is not related to modifications in the system of visual localization of the target.     

A linear memory algorithm for Baum-Welch training

Background:  

Baum-Welch training is an expectation-maximisation algorithm for training the emission and transition probabilities of hidden Markov models in a fully automated way. It can be employed as long as a training set of annotated sequences is known, and provides a rigorous way to derive parameter values which are guaranteed to be at least locally optimal. For complex hidden Markov models such as pair hidden Markov models and very long training sequences, even the most efficient algorithms for Baum-Welch training are currently too memory-consuming. This has so far effectively prevented the automatic parameter training of hidden Markov models that are currently used for biological sequence analyses.     

Spatial constancy and the brain: insights from neural networks

To form an accurate internal representation of visual space, the brain must accurately account for movements of the eyes, head or body. Updating of internal representations in response to these movements is especially important when remembering spatial information, such as the location of an object, since the brain must rely on non-visual extra-retinal signals to compensate for self-generated movements. We investigated the computations underlying spatial updating by constructing a recurrent neural network model to store and update a spatial location based on a gaze shift signal, and to do so flexibly based on a contextual cue. We observed a striking similarity between the patterns of behaviour produced by the model and monkeys trained to perform the same task, as well as between the hidden units of the model and neurons in the lateral intraparietal area (LIP). In this report, we describe the similarities between the model and single unit physiology to illustrate the usefulness of neural networks as a tool for understanding specific computations performed by the brain.