A probabilistic methodology for integrating knowledge and experiments on biological networks.

Biological systems are traditionally studied by focusing on a specific subsystem, building an intuitive model for it, and refining the model using results from carefully designed experiments. Modern experimental techniques provide massive data on the global behavior of biological systems, and systematically using these large datasets for refining existing knowledge is a major challenge. Here we introduce an extended computational framework that combines formalization of existing qualitative models, probabilistic modeling, and integration of high-throughput experimental data. Using our methods, it is possible to interpret genomewide measurements in the context of prior knowledge on the system, to assign statistical meaning to the accuracy of such knowledge, and to learn refined models with improved fit to the experiments. Our model is represented as a probabilistic factor graph, and the framework accommodates partial measurements of diverse biological elements. We study the performance of several probabilistic inference algorithms and show that hidden model variables can be reliably inferred even in the presence of feedback loops and complex logic. We show how to refine prior knowledge on combinatorial regulatory relations using hypothesis testing and derive p-values for learned model features. We test our methodology and algorithms on a simulated model and on two real yeast models. In particular, we use our method to explore uncharacterized relations among regulators in the yeast response to hyper-osmotic shock and in the yeast lysine biosynthesis system. Our integrative approach to the analysis of biological regulation is demonstrated to synergistically combine qualitative and quantitative evidence into concrete biological predictions.     

The strength of chemical linkage as a criterion for pruning metabolic graphs

MOTIVATION: A metabolic graph represents the connectivity patterns of a metabolic system, and provides a powerful framework within which the organization of metabolic reactions can be analyzed and elucidated. A common practice is to prune (i.e. remove nodes and edges) the metabolic graph prior to any analysis in order to eliminate confounding signals from the representation. Currently, this pruning process is carried out in an ad hoc fashion, resulting in discrepancies and ambiguities across studies. RESULTS: We propose a biochemically informative criterion, the strength of chemical linkage (SCL), for a systematic pruning of metabolic graphs. By analyzing the metabolic graph of Escherichia coli, we show that thresholding SCL is powerful in selecting the conventional pathways' connectivity out of the raw network connectivity when the network is restricted to the reactions collected from these pathways. Further, we argue that the root of ambiguity in pruning metabolic graphs is in the continuity of the amount of chemical content that can be conserved in reaction transformation patterns. Finally, we demonstrate how biochemical pathways can be inferred efficiently if the search procedure is guided by SCL.     

Structural distance and evolutionary relationship of networks

Exploring common features and universal qualities shared by a particular class of networks in biological and other domains is one of the important aspects of evolutionary study. In an evolving system, evolutionary mechanism can cause functional changes that forces the system to adapt to new configurations of interaction pattern between the components of that system (e.g. gene duplication and mutation play a vital role for changing the connectivity structure in many biological networks. The evolutionary relation between two systems can be retraced by their structural differences). The eigenvalues of the normalized graph Laplacian not only capture the global properties of a network, but also local structures that are produced by graph evolutions (like motif duplication or joining). The spectrum of this operator carries many qualitative aspects of a graph. Given two networks of different sizes, we propose a method to quantify the topological distance between them based on the contrasting spectrum of normalized graph Laplacian. We find that network architectures are more similar within the same class compared to between classes. We also show that the evolutionary relationships can be retraced by the structural differences using our method. We analyze 43 metabolic networks from different species and mark the prominent separation of three groups: Bacteria, Archaea and Eukarya. This phenomenon is well captured in our findings that support the other cladistic results based on gene content and ribosomal RNA sequences. Our measure to quantify the structural distance between two networks is useful to elucidate evolutionary relationships.     

Detecting and Removing Inconsistencies between Experimental Data and Signaling Network Topologies Using Integer Linear Programming on Interaction Graphs

Cross-referencing experimental data with our current knowledge of signaling network topologies is one central goal of mathematical modeling of cellular signal transduction networks. We present a new methodology for data-driven interrogation and training of signaling networks. While most published methods for signaling network inference operate on Bayesian, Boolean, or ODE models, our approach uses integer linear programming (ILP) on interaction graphs to encode constraints on the qualitative behavior of the nodes. These constraints are posed by the network topology and their formulation as ILP allows us to predict the possible qualitative changes (up, down, no effect) of the activation levels of the nodes for a given stimulus. We provide four basic operations to detect and remove inconsistencies between measurements and predicted behavior: (i) find a topology-consistent explanation for responses of signaling nodes measured in a stimulus-response experiment (if none exists, find the closest explanation); (ii) determine a minimal set of nodes that need to be corrected to make an inconsistent scenario consistent; (iii) determine the optimal subgraph of the given network topology which can best reflect measurements from a set of experimental scenarios; (iv) find possibly missing edges that would improve the consistency of the graph with respect to a set of experimental scenarios the most. We demonstrate the applicability of the proposed approach by interrogating a manually curated interaction graph model of EGFR/ErbB signaling against a library of high-throughput phosphoproteomic data measured in primary hepatocytes. Our methods detect interactions that are likely to be inactive in hepatocytes and provide suggestions for new interactions that, if included, would significantly improve the goodness of fit. Our framework is highly flexible and the underlying model requires only easily accessible biological knowledge. All related algorithms were implemented in a freely available toolbox SigNetTrainer making it an appealing approach for various applications.     

Improving gene network inference by comparing expression time-series across species, developmental stages or tissues

We present a method for gene network inference and revision based on time-series data. Gene networks are modeled using linear differential equations and a generalized stepwise multiple linear regression procedure is used to recover the interaction coefficients. Our system is designed for the recovery of gene interactions concurrently in many gene regulatory networks related by a tree or a more general graph. We show how this comparative framework can facilitate the recovery of the networks and improve the quality of the solutions inferred.     

Global dynamics of microbial communities emerge from local interaction rules

Most microbes live in spatially structured communities (e.g., biofilms) in which they interact with their neighbors through the local exchange of diffusible molecules. To understand the functioning of these communities, it is essential to uncover how these local interactions shape community-level properties, such as the community composition, spatial arrangement, and growth rate. Here, we present a mathematical framework to derive community-level properties from the molecular mechanisms underlying the cell-cell interactions for systems consisting of two cell types. Our framework consists of two parts: a biophysical model to derive the local interaction rules (i.e. interaction range and strength) from the molecular parameters underlying the cell-cell interactions and a graph based model to derive the equilibrium properties of the community (i.e. composition, spatial arrangement, and growth rate) from these local interaction rules. Our framework shows that key molecular parameters underlying the cell-cell interactions (e.g., the uptake and leakage rates of molecules) determine community-level properties. We apply our model to mutualistic cross-feeding communities and show that spatial structure can be detrimental for these communities. Moreover, our model can qualitatively recapitulate the properties of an experimental microbial community. Our framework can be extended to a variety of systems of two interacting cell types, within and beyond the microbial world, and contributes to our understanding of how community-level properties emerge from microscopic interactions between cells.     

The Logic of EGFR/ErbB Signaling: Theoretical Properties and Analysis of High-Throughput Data

The epidermal growth factor receptor (EGFR) signaling pathway is probably the best-studied receptor system in mammalian cells, and it also has become a popular example for employing mathematical modeling to cellular signaling networks. Dynamic models have the highest explanatory and predictive potential; however, the lack of kinetic information restricts current models of EGFR signaling to smaller sub-networks. This work aims to provide a large-scale qualitative model that comprises the main and also the side routes of EGFR/ErbB signaling and that still enables one to derive important functional properties and predictions. Using a recently introduced logical modeling framework, we first examined general topological properties and the qualitative stimulus-response behavior of the network. With species equivalence classes, we introduce a new technique for logical networks that reveals sets of nodes strongly coupled in their behavior. We also analyzed a model variant which explicitly accounts for uncertainties regarding the logical combination of signals in the model. The predictive power of this model is still high, indicating highly redundant sub-structures in the network. Finally, one key advance of this work is the introduction of new techniques for assessing high-throughput data with logical models (and their underlying interaction graph). By employing these techniques for phospho-proteomic data from primary hepatocytes and the HepG2 cell line, we demonstrate that our approach enables one to uncover inconsistencies between experimental results and our current qualitative knowledge and to generate new hypotheses and conclusions. Our results strongly suggest that the Rac/Cdc42 induced p38 and JNK cascades are independent of PI3K in both primary hepatocytes and HepG2. Furthermore, we detected that the activation of JNK in response to neuregulin follows a PI3K-dependent signaling pathway.     

Pairwise alignment of protein interaction networks.

With an ever-increasing amount of available data on protein-protein interaction (PPI) networks and research revealing that these networks evolve at a modular level, discovery of conserved patterns in these networks becomes an important problem. Although available data on protein-protein interactions is currently limited, recently developed algorithms have been shown to convey novel biological insights through employment of elegant mathematical models. The main challenge in aligning PPI networks is to define a graph theoretical measure of similarity between graph structures that captures underlying biological phenomena accurately. In this respect, modeling of conservation and divergence of interactions, as well as the interpretation of resulting alignments, are important design parameters. In this paper, we develop a framework for comprehensive alignment of PPI networks, which is inspired by duplication/divergence models that focus on understanding the evolution of protein interactions. We propose a mathematical model that extends the concepts of match, mismatch, and gap in sequence alignment to that of match, mismatch, and duplication in network alignment and evaluates similarity between graph structures through a scoring function that accounts for evolutionary events. By relying on evolutionary models, the proposed framework facilitates interpretation of resulting alignments in terms of not only conservation but also divergence of modularity in PPI networks. Furthermore, as in the case of sequence alignment, our model allows flexibility in adjusting parameters to quantify underlying evolutionary relationships. Based on the proposed model, we formulate PPI network alignment as an optimization problem and present fast algorithms to solve this problem. Detailed experimental results from an implementation of the proposed framework show that our algorithm is able to discover conserved interaction patterns very effectively, in terms of both accuracies and computational cost.     

A novel analytical method for evolutionary graph theory problems

Evolutionary graph theory studies the evolutionary dynamics of populations structured on graphs. A central problem is determining the probability that a small number of mutants overtake a population. Currently, Monte Carlo simulations are used for estimating such fixation probabilities on general directed graphs, since no good analytical methods exist. In this paper, we introduce a novel deterministic framework for computing fixation probabilities for strongly connected, directed, weighted evolutionary graphs under neutral drift. We show how this framework can also be used to calculate the expected number of mutants at a given time step (even if we relax the assumption that the graph is strongly connected), how it can extend to other related models (e.g. voter model), how our framework can provide non-trivial bounds for fixation probability in the case of an advantageous mutant, and how it can be used to find a non-trivial lower bound on the mean time to fixation. We provide various experimental results determining fixation probabilities and expected number of mutants on different graphs. Among these, we show that our method consistently outperforms Monte Carlo simulations in speed by several orders of magnitude. Finally we show how our approach can provide insight into synaptic competition in neurology.     

Detecting conserved interaction patterns in biological networks.

Molecular interaction data plays an important role in understanding biological processes at a modular level by providing a framework for understanding cellular organization, functional hierarchy, and evolutionary conservation. As the quality and quantity of network and interaction data increases rapidly, the problem of effectively analyzing this data becomes significant. Graph theoretic formalisms, commonly used for these analysis tasks, often lead to computationally hard problems due to their relation to subgraph isomorphism. This paper presents an innovative new algorithm, MULE, for detecting frequently occurring patterns and modules in biological networks. Using an innovative graph simplification technique based on ortholog contraction, which is ideally suited to biological networks, our algorithm renders these problems computationally tractable and scalable to large numbers of networks. We show, experimentally, that our algorithm can extract frequently occurring patterns in metabolic pathways and protein interaction networks from the KEGG, DIP, and BIND databases within seconds. When compared to existing approaches, our graph simplification technique can be viewed either as a pruning heuristic, or a closely related, but computationally simpler task. When used as a pruning heuristic, we show that our technique reduces effective graph sizes significantly, accelerating existing techniques by several orders of magnitude! Indeed, for most of the test cases, existing techniques could not even be applied without our pruning step. When used as a stand-alone analysis technique, MULE is shown to convey significant biological insights at near-interactive rates. The software, sample input graphs, and detailed results for comprehensive analysis of nine eukaryotic PPI networks are available at www.cs.purdue.edu/homes/koyuturk/mule.     

Bringing metabolic networks to life: integration of kinetic, metabolic, and proteomic data

Background

Translating a known metabolic network into a dynamic model requires reasonable guesses of all enzyme parameters. In Bayesian parameter estimation, model parameters are described by a posterior probability distribution, which scores the potential parameter sets, showing how well each of them agrees with the data and with the prior assumptions made.

Results

We compute posterior distributions of kinetic parameters within a Bayesian framework, based on integration of kinetic, thermodynamic, metabolic, and proteomic data. The structure of the metabolic system (i.e., stoichiometries and enzyme regulation) needs to be known, and the reactions are modelled by convenience kinetics with thermodynamically independent parameters. The parameter posterior is computed in two separate steps: a first posterior summarises the available data on enzyme kinetic parameters; an improved second posterior is obtained by integrating metabolic fluxes, concentrations, and enzyme concentrations for one or more steady states. The data can be heterogenous, incomplete, and uncertain, and the posterior is approximated by a multivariate log-normal distribution. We apply the method to a model of the threonine synthesis pathway: the integration of metabolic data has little effect on the marginal posterior distributions of individual model parameters. Nevertheless, it leads to strong correlations between the parameters in the joint posterior distribution, which greatly improve the model predictions by the following Monte-Carlo simulations.

Conclusion

We present a standardised method to translate metabolic networks into dynamic models. To determine the model parameters, evidence from various experimental data is combined and weighted using Bayesian parameter estimation. The resulting posterior parameter distribution describes a statistical ensemble of parameter sets; the parameter variances and correlations can account for missing knowledge, measurement uncertainties, or biological variability. The posterior distribution can be used to sample model instances and to obtain probabilistic statements about the model's dynamic behaviour.     

From classical genetics to quantitative genetics to systems biology: modeling epistasis

Gene expression data has been used in lieu of phenotype in both classical and quantitative genetic settings. These two disciplines have separate approaches to measuring and interpreting epistasis, which is the interaction between alleles at different loci. We propose a framework for estimating and interpreting epistasis from a classical experiment that combines the strengths of each approach. A regression analysis step accommodates the quantitative nature of expression measurements by estimating the effect of gene deletions plus any interaction. Effects are selected by significance such that a reduced model describes each expression trait. We show how the resulting models correspond to specific hierarchical relationships between two regulator genes and a target gene. These relationships are the basic units of genetic pathways and genomic system diagrams. Our approach can be extended to analyze data from a variety of experiments, multiple loci, and multiple environments.     

Anticipatory Functions, Digital-Analog Forms and Biosemiotics: Integrating the Tools to Model Information and Normativity in Autonomous Biological Agents

We argue that living systems process information such that functionality emerges in them on a continuous basis. We then provide a framework that can explain and model the normativity of biological functionality. In addition we offer an explanation of the anticipatory nature of functionality within our overall approach. We adopt a Peircean approach to Biosemiotics, and a dynamical approach to Digital-Analog relations and to the interplay between different levels of functionality in autonomous systems, taking an integrative approach. We then apply the underlying biosemiotic logic to a particular biological system, giving a model of the B-Cell Receptor signaling system, in order to demonstrate how biosemiotic concepts can be used to build an account of biological information and functionality. Next we show how this framework can be used to explain and model more complex aspects of biological normativity, for example, how cross-talk between different signaling pathways can be avoided. Overall, we describe an integrated theoretical framework for the emergence of normative functions and, consequently, for the way information is transduced across several interconnected organizational levels in an autonomous system, and we demonstrate how this can be applied in real biological phenomena. Our aim is to open the way towards realistic tools for the modeling of information and normativity in autonomous biological agents.     

Dynamic generation and qualitative analysis of metabolic pathways by a joint database/graph theoretical approach.

The dynamic generation and qualitative analysis of metabolic networks relying on continuously growing qualified metabolic data by a joint database/graph theoretical approach is described. The procedure is applied to analyze the connectivity of a metabolic network after enzyme removal and to subsequently perform shortest path analyses. The focus lies on the analysis of the connectivity of the metabolic network depending on model assumptions. Here we analyze the influence of the number of strongly connected components on the assignment of reversibility or irreversibility of the biochemical reactions.     

How plants changed the world: Using qualitative reasoning to explain plant macroevolution's effect on the long-term carbon cycle

We present a qualitative reasoning model of how plant colonization of land during the mid Paleozoic era (450–300 million years ago) altered the long-term carbon cycle resulting in a dramatic decrease in global atmospheric carbon dioxide levels. This model is aimed at facilitating learning and communication about how interactions between biological and geological processes drove system behavior. The model is developed in three submodels of the main system components, namely how competition for limited land habitat drove natural selection for increasing adaptations to life on land; how these adaptations resulted in increased formation of organic-rich sedimentary rocks (coal); and how these adaptations altered weathering of calcium and magnesium silicate rocks, resulting in increased deposition of inorganic carbonates in oceans. These separate submodels are then assembled to derive the full dynamic model of plant macroevolution, colonization of land, and plummeting carbon dioxide levels that occurred during the mid Paleozoic. The qualitative reasoning framework supports explicit representation of causal feedbacks — as with previously developed systems analysis models — but also supports simulation of system dynamics arising from the configuration of entities in the system. The ability of qualitative reasoning to provide causal accounts (explanations) of why certain phenomena occurred and when, is a powerful advantage over numerical simulation such as the complex GEOCARB models, where explanation must be left to interpretation by experts.     

A unified approach to joint modeling of multiple quantitative and qualitative traits in gene mapping

Using graph theory, we present a theoretical basis for mapping oligogenes in the joint presence of multiple phenotypic measurements of both quantitative and qualitative types. Various statistical models proposed earlier for several traits of solely single type are special cases of the unified approach given here. Our emphasis is on the generality of the framework, without specifying explicit assumptions about a sampling design. When information about environmental factors potentially affecting the traits is available, it can be incorporated into the genetic model. We adopt the Bayesian inferential machinery due to its firm theoretical basis and its capability of handling uncertain quantities; such as unobserved model parameters, missing marker data, and even different putative genetic models, probabilistically within a single framework. It is shown here that biological hypotheses about single gene affecting simultaneously multiple traits (pleiotropy) can be intuitively imposed as parameter constraints, leading to pleiotropic models for which posterior probabilities can be calculated. Outline of the possible implementation of the Bayesian method is described using the general reversible-jump Markov chain Monte Carlo algorithm. Some future challenges and extensions are also discussed.     

Effects of Multiple Occupancy and Interparticle Interactions on Selective Transport through Narrow Channels: Theory versus Experiment

Many biological and artificial transport channels function without direct input of metabolic energy during a transport event and without structural rearrangements involving transitions from a closed to an open state. Nevertheless, such channels are able to maintain efficient and selective transport. It has been proposed that attractive interactions between the transported molecules and the channel can increase the transport efficiency and that the selectivity of such channels can be based on the strength of the interaction of the specifically transported molecules with the channel. Herein, we study the transport through narrow channels in a framework of a general kinetic theory, which naturally incorporates multiparticle occupancy of the channel and non-single-file transport. We study how the transport efficiency and the probability of translocation through the channel are affected by interparticle interactions in the confined space inside the channel, and establish conditions for selective transport. We compare the predictions of the model with the available experimental data and find good semiquantitative agreement. Finally, we discuss applications of the theory to the design of artificial nanomolecular sieves.