Discrete Logic Modelling Optimization to Contextualize Prior Knowledge Networks Using PRUNET

Automatic mechanistic inference from large families of Boolean models generated by Monte Carlo tree search

Many important processes in biology, such as signaling and gene regulation, can be described using logic models. These logic models are typically built to behaviorally emulate experimentally observed phenotypes, which are assumed to be steady states of a biological system. Most models are built by hand and therefore researchers are only able to consider one or perhaps a few potential mechanisms. We present a method to automatically synthesize Boolean logic models with a specified set of steady states. Our method, called MC-Boomer, is based on Monte Carlo Tree Search an efficient, parallel search method using reinforcement learning. Our approach enables users to constrain the model search space using prior knowledge or biochemical interaction databases, thus leading to generation of biologically plausible mechanistic hypotheses. Our approach can generate very large numbers of data-consistent models. To help develop mechanistic insight from these models, we developed analytical tools for multi-model inference and model selection. These tools reveal the key sets of interactions that govern the behavior of the models. We demonstrate that MC-Boomer works well at reconstructing randomly generated models. Then, using single time point measurements and reasonable biological constraints, our method generates hundreds of thousands of candidate models that match experimentally validated in-vivo behaviors of the Drosophila segment polarity network. Finally we outline how our multi-model analysis procedures elucidate potentially novel biological mechanisms and provide opportunities for model-driven experimental validation.

Inferring gene regulatory networks using transcriptional profiles as dynamical attractors

Genetic regulatory networks (GRNs) regulate the flow of genetic information from the genome to expressed messenger RNAs (mRNAs) and thus are critical to controlling the phenotypic characteristics of cells. Numerous methods exist for profiling mRNA transcript levels and identifying protein-DNA binding interactions at the genome-wide scale. These enable researchers to determine the structure and output of transcriptional regulatory networks, but uncovering the complete structure and regulatory logic of GRNs remains a challenge. The field of GRN inference aims to meet this challenge using computational modeling to derive the structure and logic of GRNs from experimental data and to encode this knowledge in Boolean networks, Bayesian networks, ordinary differential equation (ODE) models, or other modeling frameworks. However, most existing models do not incorporate dynamic transcriptional data since it has historically been less widely available in comparison to "static" transcriptional data. We report the development of an evolutionary algorithm-based ODE modeling approach (named EA) that integrates kinetic transcription data and the theory of attractor matching to infer GRN architecture and regulatory logic. Our method outperformed six leading GRN inference methods, none of which incorporate kinetic transcriptional data, in predicting regulatory connections among TFs when applied to a small-scale engineered synthetic GRN in Saccharomyces cerevisiae. Moreover, we demonstrate the potential of our method to predict unknown transcriptional profiles that would be produced upon genetic perturbation of the GRN governing a two-state cellular phenotypic switch in Candida albicans. We established an iterative refinement strategy to facilitate candidate selection for experimentation; the experimental results in turn provide validation or improvement for the model. In this way, our GRN inference approach can expedite the development of a sophisticated mathematical model that can accurately describe the structure and dynamics of the in vivo GRN.

SNAIL driven by a feed forward loop motif promotes TGFβinduced epithelial to mesenchymal transition

Epithelial to Mesenchymal Transition (EMT) plays an important role in tissue regeneration, embryonic development, and cancer metastasis. Several signaling pathways are known to regulate EMT, among which the modulation of TGFβ(Transforming Growth Factor-β) induced EMT is crucial in several cancer types. Several mathematical models were built to explore the role of core regulatory circuit of ZEB/miR-200, SNAIL/miR-34 double negative feedback loops in modulating TGFβinduced EMT. Different emergent behavior including tristability, irreversible switching, existence of hybrid EMT states were inferred though these models. Some studies have explored the role of TGFβreceptor activation, SMADs nucleocytoplasmic shuttling and complex formation. Recent experiments have revealed that MDM2 along with SMAD complex regulates SNAIL expression driven EMT. Encouraged by this, in the present study we developed a mathematical model for p53/MDM2 dependent TGFβinduced EMT regulation. Inclusion of p53 brings in an additional mechanistic perspective in exploring the EM transition. The network formulated comprises a C1FFL moderating SNAIL expression involving MDM2 and SMAD complex, which functions as a noise filter and persistent detector. The C1FFL was also observed to operate as a coincidence detector driving the SNAIL dependent downstream signaling into phenotypic switching decision. Systems modelling and analysis of the devised network, displayed interesting dynamic behavior, systems response to various inputs stimulus, providing a better understanding of p53/MDM2 dependent TGF-βinduced Epithelial to Mesenchymal Transition.

Construction and contextualization approaches for protein-protein interaction networks

Protein-protein interaction network (PPIN) analysis is a widely used method to study the contextual role of proteins of interest, to predict novel disease genes, disease or functional modules, and to identify novel drug targets. PPIN-based analysis uses both generic and context-specific networks. Multiple contextualization methodologies have been described, such as shortest-path algorithms, neighborhood-based methods, and diffusion/propagation algorithms. This review discusses these methods, provides intuitive representations of PPIN contextualization, and also examines how the quality of such context-specific networks could be improved by considering additional sources of evidence. As a heuristic, we observe that tasks such as identifying disease genes, drug targets, and protein complexes should consider local neighborhoods, while uncovering disease mechanisms and discovering disease-pathways would gain from diffusion-based construction.

Assembling Disease Networks From Causal Interaction Resources

The development of high-throughput high-content technologies and the increased ease in their application in clinical settings has raised the expectation of an important impact of these technologies on diagnosis and personalized therapy. Patient genomic and expression profiles yield lists of genes that are mutated or whose expression is modulated in specific disease conditions. The challenge remains of extracting from these lists functional information that may help to shed light on the mechanisms that are perturbed in the disease, thus setting a rational framework that may help clinical decisions. Network approaches are playing an increasing role in the organization and interpretation of patients' data. Biological networks are generated by connecting genes or gene products according to experimental evidence that demonstrates their interactions. Till recently most approaches have relied on networks based on physical interactions between proteins. Such networks miss an important piece of information as they lack details on the functional consequences of the interactions. Over the past few years, a number of resources have started collecting causal information of the type protein A activates/inactivates protein B, in a structured format. This information may be represented as signed directed graphs where physiological and pathological signaling can be conveniently inspected. In this review we will (i) present and compare these resources and discuss the different scope in comparison with pathway resources; (ii) compare resources that explicitly capture causality in terms of data content and proteome coverage (iii) review how causal-graphs can be used to extract disease-specific Boolean networks.

In Silico Logical Modelling to Uncover Cooperative Interactions in Cancer

The multistep development of cancer involves the cooperation between multiple molecular lesions, as well as complex interactions between cancer cells and the surrounding tumour microenvironment. The search for these synergistic interactions using experimental models made tremendous contributions to our understanding of oncogenesis. Yet, these approaches remain labour-intensive and challenging. To tackle such a hurdle, an integrative, multidisciplinary effort is required. In this article, we highlight the use of logical computational models, combined with experimental validations, as an effective approach to identify cooperative mechanisms and therapeutic strategies in the context of cancer biology. In silico models overcome limitations of reductionist approaches by capturing tumour complexity and by generating powerful testable hypotheses. We review representative examples of logical models reported in the literature and their validation. We then provide further analyses of our logical model of Epithelium to Mesenchymal Transition (EMT), searching for additional cooperative interactions involving inputs from the tumour microenvironment and gain of function mutations in NOTCH.

Integrating Patient-Specific Information into Logic Models of Complex Diseases: Application to Acute Myeloid Leukemia

High throughput technologies such as deep sequencing and proteomics are increasingly becoming mainstream in clinical practice and support diagnosis and patient stratification. Developing computational models that recapitulate cell physiology and its perturbations in disease is a required step to help with the interpretation of results of high content experiments and to devise personalized treatments. As complete cell-models are difficult to achieve, given limited experimental information and insurmountable computational problems, approximate approaches should be considered. We present here a general approach to modeling complex diseases by embedding patient-specific genomics data into actionable logic models that take into account prior knowledge. We apply the strategy to acute myeloid leukemia (AML) and assemble a network of logical relationships linking most of the genes that are found frequently mutated in AML patients. We derive Boolean models from this network and we show that by priming the model with genomic data we can infer relevant patient-specific clinical features. Here we propose that the integration of literature-derived causal networks with patient-specific data should be explored to help bedside decisions.

Deciphering the Dynamic Transcriptional and Post-transcriptional Networks of Macrophages in the Healthy Heart and after Myocardial Injury

Macrophage plasticity has been studied in vitro, but transcriptional regulation upon injury is poorly understood. We generated a valuable dataset that captures transcriptional changes in the healthy heart and after myocardial injury, revealing a dynamic transcriptional landscape of macrophage activation. Partial deconvolution suggested that post-injury macrophages exhibit overlapping activation of pro-inflammatory and anti-inflammatory programs rather than aligning to canonical M1/M2 programs. Furthermore, simulated dynamics and experimental validation of a regulatory core of the underlying gene-regulatory network revealed a negative-feedback loop that limits initial inflammation via hypoxia-mediated upregulation of Il10. Our results also highlight the prominence of post-transcriptional regulation (miRNAs, mRNA decay, and lincRNAs) in attenuating the myocardial injury-induced inflammatory response. We also identified a cardiac-macrophage-specific gene signature (e.g., Egfr and Lifr) and time-specific markers for macrophage populations (e.g., Lyve1, Cd40, and Mrc1). Altogether, these data provide a core resource for deciphering the transcriptional network in cardiac macrophages in vivo.

Personalization of Logical Models With Multi-Omics Data Allows Clinical Stratification of Patients

Logical models of cancer pathways are typically built by mining the literature for relevant experimental observations. They are usually generic as they apply for large cohorts of individuals. As a consequence, they generally do not capture the heterogeneity of patient tumors and their therapeutic responses. We present here a novel framework, referred to as PROFILE, to tailor logical models to a particular biological sample such as a patient tumor. This methodology permits to compare the model simulations to individual clinical data, i.e., survival time. Our approach focuses on integrating mutation data, copy number alterations (CNA), and expression data (transcriptomics or proteomics) to logical models. These data need first to be either binarized or set between 0 and 1, and can then be incorporated in the logical model by modifying the activity of the node, the initial conditions or the state transition rates. The use of MaBoSS, a tool based on Monte-Carlo kinetic algorithm to perform stochastic simulations on logical models results in model state probabilities, and allows for a semi-quantitative study of the model phenotypes and perturbations. As a proof of concept, we use a published generic model of cancer signaling pathways and molecular data from METABRIC breast cancer patients. For this example, we test several combinations of data incorporation and discuss that, with these data, the most comprehensive patient-specific cancer models are obtained by modifying the nodes' activity of the model with mutations, in combination or not with CNA data, and altering the transition rates with RNA expression. We conclude that these model simulations show good correlation with clinical data such as patients' Nottingham prognostic index (NPI) subgrouping and survival time. We observe that two highly relevant cancer phenotypes derived from personalized models, Proliferation and Apoptosis, are biologically consistent prognostic factors: patients with both high proliferation and low apoptosis have the worst survival rate, and conversely. Our approach aims to combine the mechanistic insights of logical modeling with multi-omics data integration to provide patient-relevant models. This work leads to the use of logical modeling for precision medicine and will eventually facilitate the choice of patient-specific drug treatments by physicians.

Computational discovery of dynamic cell line specific Boolean networks from multiplex time-course data

Protein signaling networks are static views of dynamic processes where proteins go through many biochemical modifications such as ubiquitination and phosphorylation to propagate signals that regulate cells and can act as feed-back systems. Understanding the precise mechanisms underlying protein interactions can elucidate how signaling and cell cycle progression occur within cells in different diseases such as cancer. Large-scale protein signaling networks contain an important number of experimentally verified protein relations but lack the capability to predict the outcomes of the system, and therefore to be trained with respect to experimental measurements. Boolean Networks (BNs) are a simple yet powerful framework to study and model the dynamics of the protein signaling networks. While many BN approaches exist to model biological systems, they focus mainly on system properties, and few exist to integrate experimental data in them. In this work, we show an application of a method conceived to integrate time series phosphoproteomic data into protein signaling networks. We use a large-scale real case study from the HPN-DREAM Breast Cancer challenge. Our efficient and parameter-free method combines logic programming and model-checking to infer a family of BNs from multiple perturbation time series data of four breast cancer cell lines given a prior protein signaling network. Because each predicted BN family is cell line specific, our method highlights commonalities and discrepancies between the four cell lines. Our models have a Root Mean Square Error (RMSE) of 0.31 with respect to the testing data, while the best performant method of this HPN-DREAM challenge had a RMSE of 0.47. To further validate our results, BNs are compared with the canonical mTOR pathway showing a comparable AUROC score (0.77) to the top performing HPN-DREAM teams. In addition, our approach can also be used as a complementary method to identify erroneous experiments. These results prove our methodology as an efficient dynamic model discovery method in multiple perturbation time course experimental data of large-scale signaling networks. The software and data are publicly available at https://github.com/misbahch6/caspo-ts.

Traditional canonical signaling pathways help to understand overall signaling processes inside the cell. Large scale phosphoproteomic data provide insight into alterations among different proteins under different experimental settings. Our goal is to combine the traditional signaling networks with complex phosphoproteomic time-series data in order to unravel cell specific signaling networks. In this study, we have applied the caspo time series (caspo-ts) approach which is a combination of logic programming and model checking, over the time series phosphoproteomic dataset of the HPN-DREAM challenge to learn cell specific BNs. The learned BNs can be used to identify the cell specific topology. Our analysis suggests that caspo-ts scales to real datasets, outputting networks that are not random with a lower fitness error than the models used by the 178 methods which participated in the HPN-DREAM challenge. On the biological side, we identified the cell specific and common mechanisms (logical gates) of the cell lines.

Expert curation for building network-based dynamical models: a case study on atherosclerotic plaque formation

Knowledgebases play an increasingly important role in scientific research, where the expert curation of biological knowledge in forms that are amenable to computational analysis (using ontologies for example)–provides a significant added value and enables new types of computational analyses for high throughput datasets. In this work, we demonstrate how expert curation can also play a more direct role in research, by supporting the use of network-based dynamical models to study a specific biological process. This curation effort is focused on the regulatory interactions between biological entities, such as genes or proteins and compounds, which may interact with each other in a complex manner, including regulatory complexes and conditional dependencies between co-regulators. This critical information has to be captured and encoded in a computable manner, which is currently far beyond the current capabilities of automatically constructed network. As a case study, we report here the prior knowledge network constructed by the sysVASC consortium to model the biological events leading to the formation of atherosclerotic plaques, during the onset of cardiovascular disease and discuss some specific examples to illustrate the main pitfalls and added value provided by the expert curation during this endeavor.

Database URL: http://biomodels.caltech.edu

Boolean regulatory network reconstruction using literature based knowledge with a genetic algorithm optimization method

Background

Prior knowledge networks (PKNs) provide a framework for the development of computational biological models, including Boolean models of regulatory networks which are the focus of this work. PKNs are created by a painstaking process of literature curation, and generally describe all relevant regulatory interactions identified using a variety of experimental conditions and systems, such as specific cell types or tissues. Certain of these regulatory interactions may not occur in all biological contexts of interest, and their presence may dramatically change the dynamical behaviour of the resulting computational model, hindering the elucidation of the underlying mechanisms and reducing the usefulness of model predictions. Methods are therefore required to generate optimized contextual network models from generic PKNs.

Results

We developed a new approach to generate and optimize Boolean networks, based on a given PKN. Using a genetic algorithm, a model network is built as a sub-network of the PKN and trained against experimental data to reproduce the experimentally observed behaviour in terms of attractors and the transitions that occur between them under specific perturbations. The resulting model network is therefore contextualized to the experimental conditions and constitutes a dynamical Boolean model closer to the observed biological process used to train the model than the original PKN. Such a model can then be interrogated to simulate response under perturbation, to detect stable states and their properties, to get insights into the underlying mechanisms and to generate new testable hypotheses.

Conclusions

Generic PKNs attempt to synthesize knowledge of all interactions occurring in a biological process of interest, irrespective of the specific biological context. This limits their usefulness as a basis for the development of context-specific, predictive dynamical Boolean models. The optimization method presented in this article produces specific, contextualized models from generic PKNs. These contextualized models have improved utility for hypothesis generation and experimental design. The general applicability of this methodological approach makes it suitable for a variety of biological systems and of general interest for biological and medical research. Our method was implemented in the software optimusqual, available online at http://www.vital-it.ch/software/optimusqual/.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-016-1287-z) contains supplementary material, which is available to authorized users.

Reconstruction and Application of Protein-Protein Interaction Network

The protein-protein interaction network (PIN) is a useful tool for systematic investigation of the complex biological activities in the cell. With the increasing interests on the proteome-wide interaction networks, PINs have been reconstructed for many species, including virus, bacteria, plants, animals, and humans. With the development of biological techniques, the reconstruction methods of PIN are further improved. PIN has gradually penetrated many fields in biological research. In this work we systematically reviewed the development of PIN in the past fifteen years, with respect to its reconstruction and application of function annotation, subsystem investigation, evolution analysis, hub protein analysis, and regulation mechanism analysis. Due to the significant role of PIN in the in-depth exploration of biological process mechanisms, PIN will be preferred by more and more researchers for the systematic study of the protein systems in various kinds of organisms.

Gene Regulatory Network Inference of Immunoresponsive Gene 1 (IRG1) Identifies Interferon Regulatory Factor 1 (IRF1) as Its Transcriptional Regulator in Mammalian Macrophages

Immunoresponsive gene 1 (IRG1) is one of the highest induced genes in macrophages under pro-inflammatory conditions. Its function has been recently described: it codes for immune-responsive gene 1 protein/cis-aconitic acid decarboxylase (IRG1/CAD), an enzyme catalysing the production of itaconic acid from cis-aconitic acid, a tricarboxylic acid (TCA) cycle intermediate. Itaconic acid possesses specific antimicrobial properties inhibiting isocitrate lyase, the first enzyme of the glyoxylate shunt, an anaplerotic pathway that bypasses the TCA cycle and enables bacteria to survive on limited carbon conditions. To elucidate the mechanisms underlying itaconic acid production through IRG1 induction in macrophages, we examined the transcriptional regulation of IRG1. To this end, we studied IRG1 expression in human immune cells under different inflammatory stimuli, such as TNFα and IFNγ, in addition to lipopolysaccharides. Under these conditions, as previously shown in mouse macrophages, IRG1/CAD accumulates in mitochondria. Furthermore, using literature information and transcription factor prediction models, we re-constructed raw gene regulatory networks (GRNs) for IRG1 in mouse and human macrophages. We further implemented a contextualization algorithm that relies on genome-wide gene expression data to infer putative cell type-specific gene regulatory interactions in mouse and human macrophages, which allowed us to predict potential transcriptional regulators of IRG1. Among the computationally identified regulators, siRNA-mediated gene silencing of interferon regulatory factor 1 (IRF1) in macrophages significantly decreased the expression of IRG1/CAD at the gene and protein level, which correlated with a reduced production of itaconic acid. Using a synergistic approach of both computational and experimental methods, we here shed more light on the transcriptional machinery of IRG1 expression and could pave the way to therapeutic approaches targeting itaconic acid levels.