Benjamin M. Gyori

OpenComet: an automated tool for comet assay image analysis.

Reactive species such as free radicals are constantly generated in vivo and DNA is the most important target of oxidative stress. Oxidative DNA damage is used as a predictive biomarker to monitor the risk of development of many diseases. The comet assay is widely used for measuring oxidative DNA damage at a single cell level. The analysis of comet assay output images, however, poses considerable challenges. Commercial software is costly and restrictive, while free software generally requires laborious manual tagging of cells. This paper presents OpenComet, an open-source software tool providing automated analysis of comet assay images. It uses a novel and robust method for finding comets based on geometric shape attributes and segmenting the comet heads through image intensity profile analysis. Due to automation, OpenComet is more accurate, less prone to human bias, and faster than manual analysis. A live analysis functionality also allows users to analyze images captured directly from a microscope. We have validated OpenComet on both alkaline and neutral comet assay images as well as sample images from existing software packages. Our results show that OpenComet achieves high accuracy with significantly reduced analysis time.

Statistical Model Checking Based Calibration and Analysis of Bio-pathway Models

We present a statistical model checking SMC based framework for studying ordinary differential equation ODE models of bio-pathways. We address cell-to-cell variability explicitly by using probability distributions to model initial concentrations and kinetic rate values. This implicitly defines a distribution over a set of ODE trajectories, the properties of which are to be characterized. The core component of our framework is an SMC procedure for verifying the dynamical properties of an ODE system accompanied by such prior distributions. To cope with the imprecise nature of biological data, we use a formal specification logic that allows us to encode both qualitative properties and experimental data. Using SMC, we verify such specifications in a tractable way, independent of the system size. This further enables us to develop SMC based parameter estimation and sensitivity analysis procedures. We have evaluated our method on two large pathway models, namely, the segmentation clock network and the thrombin-dependent MLC phosphorylation pathway. The results show that our method scales well and yields good parameter estimates that are robust. Our sensitivity analysis framework leads to interesting insights about the underlying dynamics of these systems.

From word models to executable models of signaling networks using automated assembly

Word models (natural language descriptions of molecular mechanisms) are a common currency in spoken and written communication in biomedicine but are of limited use in predicting the behavior of complex biological networks. We present an approach to building computational models directly from natural language using automated assembly. Molecular mechanisms described in simple English are read by natural language processing algorithms, converted into an intermediate representation, and assembled into executable or network models. We have implemented this approach in the Integrated Network and Dynamical Reasoning Assembler (INDRA), which draws on existing natural language processing systems as well as pathway information in Pathway Commons and other online resources. We demonstrate the use of INDRA and natural language to model three biological processes of increasing scope: (i) p53 dynamics in response to DNA damage, (ii) adaptive drug resistance in BRAF‐V600E‐mutant melanomas, and (iii) the RAS signaling pathway. The use of natural language makes the task of developing a model more efficient and it increases model transparency, thereby promoting collaboration with the broader biology community.

Hypothesis testing for Markov chain Monte Carlo

Testing between hypotheses, when independent sampling is possible, is a well developed subject. In this paper, we propose hypothesis tests that are applicable when the samples are obtained using Markov chain Monte Carlo. These tests are useful when one is interested in deciding whether the expected value of a certain quantity is above or below a given threshold. We show non-asymptotic error bounds and bounds on the expected number of samples for three types of tests, a fixed sample size test, a sequential test with indifference region, and a sequential test without indifference region. Our tests can lead to significant savings in sample size. We illustrate our results on an example of Bayesian parameter inference involving an ODE model of a biochemical pathway.

Approximate Probabilistic Verification of Hybrid Systems

Hybrid systems whose mode dynamics are governed by non-linear ordinary differential equations (ODEs) are often a natural model for biological processes. However such models are difficult to analyze. To address this, we develop a probabilistic analysis method by approximating the mode transitions as stochastic events. We assume that the probability of making a mode transition is proportional to the measure of the set of pairs of time points and value states at which the mode transition is enabled. To ensure a sound mathematical basis, we impose a natural continuity property on the non-linear ODEs. We also assume that the states of the system are observed at discrete time points but that the mode transitions may take place at any time between two successive discrete time points. This leads to a discrete time Markov chain as a probabilistic approximation of the hybrid system. We then show that for BLTL (bounded linear time temporal logic) specifications the hybrid system meets a specification iff its Markov chain approximation meets the same specification with probability 1. Based on this, we formulate a sequential hypothesis testing procedure for verifying–approximately–that the Markov chain meets a BLTL specification with high probability. Our case studies on cardiac cell dynamics and the circadian rhythm indicate that our scheme can be applied in a number of realistic settings.

Parallelized Parameter Estimation of Biological Pathway Models

We develop a GPU based technique to analyze bio-pathway models consisting of systems of ordinary differential equations (ODEs). A key component in our technique is an online procedure for verifying whether a numerically generated trajectory of a model satisfies a property expressed in bounded linear temporal logic. Using this procedure, we construct a statistical model checking algorithm which exploits the massive parallelism offered by GPUs while respecting the severe constraints imposed by their memory hierarchy and the hardware execution model. To demonstrate the computational power of our method, we use it to solve the parameter estimation problem for bio-pathway models. With three realistic benchmarks, we show that the proposed technique is computationally efficient and scales well with the number of GPU units deployed. Since both the verification framework and the computational platform are generic, our scheme can be used to solve a variety of analysis problems for models consisting of large systems of ODEs.

Encoding growth factor identity in the temporal dynamics of FoxO3 under the combinatorial control of ERK and AKT Kinases

Extracellular growth factors signal to transcription factors via a limited number of cytoplasmic kinase cascades. It remains unclear how such cascades encode ligand identities and concentrations. In this paper, we use live-cell imaging and statistical modeling to study FOXO3, a transcription factor regulating diverse aspects of cellular physiology that is under combinatorial control. We show that FOXO3 nuclear-to-cytosolic translocation has two temporally distinct phases varying in magnitude with growth factor identity and cell type. These phases comprise synchronous translocation soon after ligand addition followed by an extended back-and-forth shuttling; this shuttling is pulsatile and does not have a characteristic frequency, unlike a simple oscillator. Early and late dynamics are differentially regulated by Akt and ERK and have low mutual information, potentially allowing the two phases to encode different information. In cancer cells in which ERK and Akt are dysregulated by oncogenic mutation, the diversity of states is lower.

FamPlex: a resource for entity recognition and relationship resolution of human protein families and complexes in biomedical text mining

AbstractBackgroundFor automated reading of scientific publications to extract useful information about molecular mechanisms it is critical that genes, proteins and other entities be correctly associated with uniform identifiers, a process known as named entity linking or “grounding.” Correct grounding is essential for resolving relationships among mined information, curated interaction databases, and biological datasets. The accuracy of this process is largely dependent on the availability of machine-readable resources associating synonyms and abbreviations commonly found in biomedical literature with uniform identifiers.ResultsIn a task involving automated reading of ∼215,000 articles using the REACH event extraction software we found that grounding was disproportionately inaccurate for multi-protein families (e.g., “AKT”) and complexes with multiple subunits (e.g.“NF- κB”). To address this problem we constructed FamPlex, a manually curated resource defining protein families and complexes as they are commonly encountered in biomedical text. In FamPlex the gene-level constituents of families and complexes are defined in a flexible format allowing for multi-level, hierarchical membership. To create FamPlex, text strings corresponding to entities were identified empirically from literature and linked manually to uniform identifiers; these identifiers were also mapped to equivalent entries in multiple related databases. FamPlex also includes curated prefix and suffix patterns that improve named entity recognition and event extraction. Evaluation of REACH extractions on a test corpus of ∼54,000 articles showed that FamPlex significantly increased grounding accuracy for families and complexes (from 15 to 71%). The hierarchical organization of entities in FamPlex also made it possible to integrate otherwise unconnected mechanistic information across families, subfamilies, and individual proteins. Applications of FamPlex to the TRIPS/DRUM reading system and the Biocreative VI Bioentity Normalization Task dataset demonstrated the utility of FamPlex in other settings.ConclusionFamPlex is an effective resource for improving named entity recognition, grounding, and relationship resolution in automated reading of biomedical text. The content in FamPlex is available in both tabular and Open Biomedical Ontology formats at https://github.com/sorgerlab/famplex under the Creative Commons CC0 license and has been integrated into the TRIPS/DRUM and REACH reading systems.

Bioentities: a resource for entity recognition and relationship resolution in biomedical text mining

Background For automated reading of scientific publications to extract useful information about molecular mechanisms it is critical that genes, proteins and other entities be correctly associated with uniform identifiers, a process known as named entity linking or “grounding.” Correct grounding is essential for resolving relationships among mined information, curated interaction databases, and biological datasets. The accuracy of this process is largely dependent on the availability of machine-readable resources associating synonyms and abbreviations commonly found in biomedical literature with uniform identifiers. Results In a task involving automated reading of ∼215,000 articles using the REACH event extraction software we found that grounding was disproportionately inaccurate for multi-protein families (e.g., “AKT”) and complexes with multiple subunits (e.g.”NF-κB”). To address this problem we constructed FamPlex, a manually curated resource defining protein families and complexes as they are commonly encountered in biomedical text. In FamPlex the gene-level constituents of families and complexes are defined in a flexible format allowing for multi-level, hierarchical membership. To create FamPlex, text strings corresponding to entities were identified empirically from literature and linked manually to uniform identifiers; these identifiers were also mapped to equivalent entries in multiple related databases. FamPlex also includes curated prefix and suffix patterns that improve named entity recognition and event extraction. Evaluation of REACH extractions on a test corpus of ∼54,000 articles showed that FamPlex significantly increased grounding accuracy for families and complexes (from 15% to 71%). The hierarchical organization of entities in FamPlex also made it possible to integrate otherwise unconnected mechanistic information across families, subfamilies, and individual proteins. Applications of FamPlex to the TRIPS/DRUM reading system and the Biocreative VI Bioentity Normalization Task dataset demonstrated the utility of FamPlex in other settings. Conclusion FamPlex is an effective resource for improving named entity recognition, grounding, and relationship resolution in automated reading of biomedical text. The content in FamPlex is available in both tabular and Open Biomedical Ontology formats at https://github.com/sorgerlab/famplex under the Creative Commons CC0 license and has been integrated into the TRIPS/DRUM and REACH reading systems.

Simulation-Guided Parameter Synthesis for Chance-Constrained Optimization of Control Systems

We consider the problem of parameter synthesis for black-box systems whose operations are jointly influenced by a set of “tunable parameters” under the control of designers, and a set of uncontrollable stochastic parameters. The goal is to find values of the tunable parameters that ensure the satisfaction of given performance requirements with a high probability. Such problems are common in robust system design, including feedback controllers, biomedical devices, and many others. These can be naturally cast as chance-constrained optimization problems, which however, are hard to solve precisely. We present a simulation-based approach that provides a piecewise approximation of a certain quantile function for the responses of interest. Using the piecewise approximations as objective functions, a collection of local optima are estimated, from which a global search based on simulated annealing is performed. The search yields tunable parameter values at which the performance requirements are satisfied with a high probability, despite variations in the stochastic parameters. Our approach is applied to three benchmarks: an insulin infusion pump model for type-1 diabetic patients, a robust flight control problem for fixed-wing aircrafts, and an ODE-based apoptosis model from system biology.