Atrial fibrillation (AF) is a common cardiac arrhythmia that affects 1% of the population and is associated with high levels of morbidity and all-cause mortality. Catheter ablation (CA) has become one of the first line treatments for AF, but the success rates of CA and other clinical treatments remain suboptimal. The need to improve clinical outcomes warrants the optimisation of CA therapy. In this study, we develop a novel deep learning method to identify specific ablation patterns that terminate AF efficiently. To achieve this, we simulate typical AF ablation scenarios using computational models of 2D atrial tissue, and use the simulation outcomes as inputs for a deep neural network. The network is trained, validated and then applied to classify the scenarios and predict the optimal CA pattern in each scenario. For the validation dataset, the overall accuracy in identifying the best CA strategy is recorded at 79%. The study provides proof of concept that deep neural networks can learn from computational models of AF and help optimise CA therapy.

The higher-order patterning of extra-cellular matrix in normal and pathological tissues has profound consequences on tissue function. Whilst studies have documented both how fibroblasts create and maintain individual matrix fibers and how cell migration is altered by the fibers they interact with, a model unifying these two aspects of tissue organization is lacking. Here we use computational modelling to understand the effect of this interconnectivity between fibroblasts and matrix at the mesoscale level. We created a unique adaptation to the Vicsek flocking model to include feedback from a second layer representing the matrix, and use experimentation to parameterize our model and validate model-driven hypotheses. Our two-layer model demonstrates that feedback between fibroblasts and matrix increases matrix diversity creating higher-order patterns. The model can quantitatively recapitulate matrix patterns of tissues in vivo. Cells follow matrix fibers irrespective of when the matrix fibers were deposited, resulting in feedback with the matrix acting as temporal 'memory' to collective behaviour, which creates diversity in topology. We also establish conditions under which matrix can be remodelled from one pattern to another. Our model elucidates how simple rules defining fibroblast-matrix interactions are sufficient to generate complex tissue patterns.

MOTIVATION: In order to function, proteins frequently bind to one another and form 3D assemblies. Knowledge of the atomic details of these structures helps our understanding of how proteins work together, how mutations can lead to disease, and facilitates the designing of drugs which prevent or mimic the interaction. RESULTS: Atomic modeling of protein-protein interactions requires the selection of near-native structures from a set of docked poses based on their calculable properties. By considering this as an information retrieval problem, we have adapted methods developed for Internet search ranking and electoral voting into IRaPPA, a pipeline integrating biophysical properties. The approach enhances the identification of near-native structures when applied to four docking methods, resulting in a near-native appearing in the top 10 solutions for up to 50% of complexes benchmarked, and up to 70% in the top 100. AVAILABILITY AND IMPLEMENTATION: IRaPPA has been implemented in the SwarmDock server ( http://bmm.crick.ac.uk/∼SwarmDock/ ), pyDock server ( http://life.bsc.es/pid/pydockrescoring/ ) and ZDOCK server ( http://zdock.umassmed.edu/ ), with code available on request. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

We present optimisations applied to a bespoke bio-physical molecular dynamics simulation designed to investigate chromosome condensation. Our primary focus is on domain-specific algorithmic improvements to determining short-range interaction forces between particles, as certain qualities of the simulation render traditional methods less effective. We implement tuned versions of the code for both traditional CPU architectures and the modern many-core architecture found in the Intel Xeon Phi coprocessor and compare their effectiveness. We achieve speed-ups starting at a factor of 10 over the original code, facilitating more detailed and larger-scale experiments.

Abstract Summary: Protein–protein interactions are central to almost all biological functions, and the atomic details of such interactions can yield insights into the mechanisms that underlie these functions. We present a web server that wraps and extends the SwarmDock flexible protein–protein docking algorithm. After uploading PDB files of the binding partners, the server generates low energy conformations and returns a ranked list of clustered docking poses and their corresponding structures. The user can perform full global docking, or focus on particular residues that are implicated in binding. The server is validated in the CAPRI blind docking experiment, against the most current docking benchmark, and against the ClusPro docking server, the highest performing server currently available. Availability: The server is freely available and can be accessed at: http://bmm.cancerresearchuk.org/%7ESwarmDock/. Contact: [email protected] Supplementary information: Supplementary data are available at Bioinformatics online.

BACKGROUND: Protein-protein docking, which aims to predict the structure of a protein-protein complex from its unbound components, remains an unresolved challenge in structural bioinformatics. An important step is the ranking of docked poses using a scoring function, for which many methods have been developed. There is a need to explore the differences and commonalities of these methods with each other, as well as with functions developed in the fields of molecular dynamics and homology modelling. RESULTS: We present an evaluation of 115 scoring functions on an unbound docking decoy benchmark covering 118 complexes for which a near-native solution can be found, yielding top 10 success rates of up to 58%. Hierarchical clustering is performed, so as to group together functions which identify near-natives in similar subsets of complexes. Three set theoretic approaches are used to identify pairs of scoring functions capable of correctly scoring different complexes. This shows that functions in different clusters capture different aspects of binding and are likely to work together synergistically. CONCLUSIONS: All functions designed specifically for docking perform well, indicating that functions are transferable between sampling methods. We also identify promising methods from the field of homology modelling. Further, differential success rates by docking difficulty and solution quality suggest a need for flexibility-dependent scoring. Investigating pairs of scoring functions, the set theoretic measures identify known scoring strategies as well as a number of novel approaches, indicating promising augmentations of traditional scoring methods. Such augmentation and parameter combination strategies are discussed in the context of the learning-to-rank paradigm.

Predicting the effects of mutations on the kinetic rate constants of protein-protein interactions is central to both the modeling of complex diseases and the design of effective peptide drug inhibitors. However, while most studies have concentrated on the determination of association rate constants, dissociation rates have received less attention. In this work we take a novel approach by relating the changes in dissociation rates upon mutation to the energetics and architecture of hotspots and hotregions, by performing alanine scans pre- and post-mutation. From these scans, we design a set of descriptors that capture the change in hotspot energy and distribution. The method is benchmarked on 713 kinetically characterized mutations from the SKEMPI database. Our investigations show that, with the use of hotspot descriptors, energies from single-point alanine mutations may be used for the estimation of off-rate mutations to any residue type and also multi-point mutations. A number of machine learning models are built from a combination of molecular and hotspot descriptors, with the best models achieving a Pearson's Correlation Coefficient of 0.79 with experimental off-rates and a Matthew's Correlation Coefficient of 0.6 in the detection of rare stabilizing mutations. Using specialized feature selection models we identify descriptors that are highly specific and, conversely, broadly important to predicting the effects of different classes of mutations, interface regions and complexes. Our results also indicate that the distribution of the critical stability regions across protein-protein interfaces is a function of complex size more strongly than interface area. In addition, mutations at the rim are critical for the stability of small complexes, but consistently harder to characterize. The relationship between hotregion size and the dissociation rate is also investigated and, using hotspot descriptors which model cooperative effects within hotregions, we show how the contribution of hotregions of different sizes, changes under different cooperative effects.

Gauging the systemic effects of non-synonymous single nucleotide polymorphisms (nsSNPs) is an important topic in the pursuit of personalized medicine. However, it is a non-trivial task to understand how a change at the protein structure level eventually affects a cell's behavior. This is because complex information at both the protein and pathway level has to be integrated. Given that the idea of integrating both protein and pathway dynamics to estimate the systemic impact of missense mutations in proteins remains predominantly unexplored, we investigate the practicality of such an approach by formulating mathematical models and comparing them with experimental data to study missense mutations. We present two case studies: (1) interpreting systemic perturbation for mutations within the cell cycle control mechanisms (G2 to mitosis transition) for yeast; (2) phenotypic classification of neuron-related human diseases associated with mutations within the mitogen-activated protein kinase (MAPK) pathway. We show that the application of simplified mathematical models is feasible for understanding the effects of small sequence changes on cellular behavior. Furthermore, we show that the systemic impact of missense mutations can be effectively quantified as a combination of protein stability change and pathway perturbation.

The prediction of protein-protein kinetic rate constants provides a fundamental test of our understanding of molecular recognition, and will play an important role in the modeling of complex biological systems. In this paper, a feature selection and regression algorithm is applied to mine a large set of molecular descriptors and construct simple models for association and dissociation rate constants using empirical data. Using separate test data for validation, the predicted rate constants can be combined to calculate binding affinity with accuracy matching that of state of the art empirical free energy functions. The models show that the rate of association is linearly related to the proportion of unbound proteins in the bound conformational ensemble relative to the unbound conformational ensemble, indicating that the binding partners must adopt a geometry near to that of the bound prior to binding. Mirroring the conformational selection and population shift mechanism of protein binding, the models provide a strong separate line of evidence for the preponderance of this mechanism in protein-protein binding, complementing structural and theoretical studies.

MOTIVATION: Accurate binding free energy functions for protein-protein interactions are imperative for a wide range of purposes. Their construction is predicated upon ascertaining the factors that influence binding and their relative importance. A recent benchmark of binding affinities has allowed, for the first time, the evaluation and construction of binding free energy models using a diverse set of complexes, and a systematic assessment of our ability to model the energetics of conformational changes. RESULTS: We construct a large set of molecular descriptors using commonly available tools, introducing the use of energetic factors associated with conformational changes and disorder to order transitions, as well as features calculated on structural ensembles. The descriptors are used to train and test a binding free energy model using a consensus of four machine learning algorithms, whose performance constitutes a significant improvement over the other state of the art empirical free energy functions tested. The internal workings of the learners show how the descriptors are used, illuminating the determinants of protein-protein binding. AVAILABILITY: The molecular descriptor set and descriptor values for all complexes are available in the Supplementary Material. A web server for the learners and coordinates for the bound and unbound structures can be accessed from the website: http://bmm.cancerresearchuk.org/~Affinity. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Cellular structures and functions rely on a complex network of regulated protein interactions, which are further based on complex molecular binding mechanisms. Three key questions are usually asked: a) how does a protein find its partner, b) at what rate do they interact and c) how they achieve the previous 2 goals in vivo. While these questions are usually studied in isolation, we believe that unique observations will be made by carrying out holistic investigation of the three problems under one roof, bringing together molecular docking, interaction kinetics and macromolecular crowding. We have developed a computational package capable of simulating a crowded macromolecular environment, in which proteins are modelled as full-atom parameterised rigid bodies. It is for the first time that specific protein-protein interactions can be investigated in a multi-macromolecular environment where all simulated bodies bear the same atomic resolution and forcefield parameters. It is also for the first time that the association rate constants, kon, can be scored by retention time of specific encounter complexes formed during the simulation and correlate with those experimentally determined. Calibrated on the experimentally determined kon in non-crowded and crowded environments, we revealed the possible molecular binding mechanisms and predicted the influence on such mechanisms from environmental crowders, of which ambient electrostatic interactions were shown to play an important part. Figure 1 . Under one roof: through crowded simulation (1), we are able to investigate pairwise docking (2), time-course information of encounter complexes (3) and the influences to protein kinetics by macromolecular crowding (4) at the same time and in the same interaction environment. Adapted from Ref [ 1 ] and a manuscript to submit. Our simulation effectively bridges the gap between structural studies of individual macromolecular complexes and their kinetic behaviour in a physiologically crowded condition. Potentially, this offers further connection to systemic approaches of studying regulatory network kinetics, as well as the enhancement of our ability to probe, manipulate and engineer molecular interactions in a simulated, but more realistic cellular environment.

Vascular abnormalities contribute to many diseases such as cancer and diabetic retinopathy. In angiogenesis new blood vessels, headed by a migrating tip cell, sprout from pre-existing vessels in response to signals, e.g., vascular endothelial growth factor (VEGF). Tip cells meet and fuse (anastomosis) to form blood-flow supporting loops. Tip cell selection is achieved by Dll4-Notch mediated lateral inhibition resulting, under normal conditions, in an interleaved arrangement of tip and non-migrating stalk cells. Previously, we showed that the increased VEGF levels found in many diseases can cause the delayed negative feedback of lateral inhibition to produce abnormal oscillations of tip/stalk cell fates. Here we describe the development and implementation of a novel physics-based hierarchical agent model, tightly coupled to in vivo data, to explore the system dynamics as perpetual lateral inhibition combines with tip cell migration and fusion. We explore the tipping point between normal and abnormal sprouting as VEGF increases. A novel filopodia-adhesion driven migration mechanism is presented and validated against in vivo data. Due to the unique feature of ongoing lateral inhibition, 'stabilised' tip/stalk cell patterns show sensitivity to the formation of new cell-cell junctions during fusion: we predict cell fates can reverse. The fusing tip cells become inhibited and neighbouring stalk cells flip fate, recursively providing new tip cells. Junction size emerges as a key factor in establishing a stable tip/stalk pattern. Cell-cell junctions elongate as tip cells migrate, which is shown to provide positive feedback to lateral inhibition, causing it to be more susceptible to pathological oscillations. Importantly, down-regulation of the migratory pathway alone is shown to be sufficient to rescue the sprouting system from oscillation and restore stability. Thus we suggest the use of migration inhibitors as therapeutic agents for vascular normalisation in cancer.

MOTIVATION: The study of interactomes, or networks of protein-protein interactions, is increasingly providing valuable information on biological systems. Here we report a study of cancer proteins in an extensive human protein-protein interaction network constructed by computational methods. RESULTS: We show that human proteins translated from known cancer genes exhibit a network topology that is different from that of proteins not documented as being mutated in cancer. In particular, cancer proteins show an increase in the number of proteins they interact with. They also appear to participate in central hubs rather than peripheral ones, mirroring their greater centrality and participation in networks that form the backbone of the proteome. Moreover, we show that cancer proteins contain a high ratio of highly promiscuous structural domains, i.e., domains with a high propensity for mediating protein interactions. These observations indicate an underlying evolutionary distinction between the two groups of proteins, reflecting the central roles of proteins, whose mutations lead to cancer. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: The interactome data are available though the PIP (Potential Interactions of Proteins) web server at http://bmm.cancerresearchuk.org/servers/pip. Further additional material is available at http://bmm.cancerresearchuk.org/servers/pip/bioinformatics/

MOTIVATION: A wide variety of methods for the construction of an atomic model for a given amino acid sequence are known, the more accurate being those that use experimentally determined structures as templates. However, far fewer methods are aimed at refining these models. The approach presented here carefully blends models created by several different means, in an attempt to combine the good quality regions from each into a final, more refined, model. RESULTS: We describe here a number of refinement operators (collectively, 'move-set') that enable a relatively large region of conformational space to be searched. This is used within a genetic algorithm that reshuffles and repacks structural components. The utility of the move-set is demonstrated by introducing a cost function, containing both physical and other components guiding the input structures towards the target structure. We show that our move-set has the potential to improve the conformation of models and that this improvement can be beyond even the best template for some comparative modelling targets. AVAILABILITY: The populus software package and the source code are available at http://bmm.cancerresearchuk.org/~offman01/populus.html.

BACKGROUND: Protein-protein interactions have traditionally been studied on a small scale, using classical biochemical methods to investigate the proteins of interest. More recently large-scale methods, such as two-hybrid screens, have been utilised to survey extensive portions of genomes. Current high-throughput approaches have a relatively high rate of errors, whereas in-depth biochemical studies are too expensive and time-consuming to be practical for extensive studies. As a result, there are gaps in our knowledge of many key biological networks, for which computational approaches are particularly suitable. RESULTS: We constructed networks, or 'interactomes', of putative protein-protein interactions in the rat proteome--the rat being an organism extensively used for cancer studies. This was achieved by integrating experimental protein-protein interaction data from many species and translating this data into the reference frame of the rat. The putative rat protein interactions were given confidence scores based on their homology to proteins that have been experimentally observed to interact. The confidence score was furthermore weighted according to the extent of the experimental evidence, giving a higher weight to more frequently observed interactions. The scoring function was subsequently validated and networks constructed around key proteins, identified as being highly up- or down-regulated in rat cell lines of high metastatic potential. Using clustering methods on the networks, we have identified key protein communities involved in cancer metastasis. CONCLUSION: The protein network generation and subsequent network analysis used here, were shown to be useful for highlighting key proteins involved in metastasis. This approach, in conjunction with microarray expression data, can be extended to other species, thereby suggesting possible pathways around proteins of interest.

Abstract Summary: To optimize the search for structural templates in protein comparative modelling, the query sequence is split into domains. The initial list of templates for each domain, extracted from PFAM plus PDB and SCOP, is then ranked according to sequence identity (%ID), coverage and resolution. If %ID is less than 30, secondary structure matching is used to filter out false templates. Availability: http://www.bmm.icnet.uk/~3djigsaw/dom_fish Contact: [email protected]@cancer.org.uk * To whom correspondence should be addressed.