Systems biology solutions offered by CFDRC include the development of customized computational methods to enable the construction of intracellular pathway models, and associated quantitative, dynamic, predictive algorithms to analyze the resultant pathway models for use in biomarker discovery, assay design, and therapeutic target identification. A brief flavor of our capabilities is presented below.

Model Development
Pathway models consist of nodes (representing genes, proteins, or metabolites) and edges (representing the interactions between the various nodes), providing a framework for the (1) management of the knowledge base of biological information, (2) systematic interrogation and experimental verification of the biochemical network. We have developed a variety of detailed models from first-principles using high throughput data (genomic, proteomic, and metabolomic), scientific literature, and expert guidance. The pathway models we have developed to date include the response of neurons to organophosphorus compounds (nerve agents and pesticides), and a pathway model describing the response of neurons to the antimalarial mefloquine.

Optimization
One of the core areas of expertise within the group is in the area of optimization applied to pathway models. We have developed a suite of tools with capabilities that include the selection of an optimal model structure given a set of data, and advanced parameter estimation to determine an optimal parameter set for a given model structure.

Model Analysis
Once a predictive, dynamic pathway model has been developed and distilled in the form of ordinary differential equations, analysis of the system dynamics and parameter sensitivity can be quantified using the suite of solvers and tools available at CFDRC. We offer other continuum and stochastic integration methods, as well as a hybrid continuum/stochastic solver to accurately simulate the internal cellular dynamics, given that the kinetics of each reaction is known.
However, large pathway models (more than 100 species) present significant challenges for most commonly applied modeling techniques. Foremost among the problems is that very few interactions in a large pathway model have well characterized chemical kinetics, which eliminates many of the ordinary differential equation-based approaches. We have developed a novel method based on boolean pseudodynamics, known as the Boolean Network Dynamics Target Identification (BNDTI) algorithm, that a priori ranks the network nodes based on their importance in the efficient and robust operation of the network. This algorithm has been successfully applied to a priori determine the critical network interactions, and published in the

Development of Dynamic, Predictive, Pathway Models of Toxin Action

Pictorial summary of the literature concerning the neurotoxic response to OP compounds.

Neuropathic effects associated with long-term exposure to sub-lethal levels of OPCs (nerve agents, pesticides) are serious health concerns for soldiers and civilians. The goal of this research is to identify exposure biomarkers and intervention targets for OPC-induced neurotoxicity, using a systems biology-centered methodology. Our approach combines temporal measurement of cellular response with detailed computational analysis of response pathways. Temporal profiles of PC12 response to OPC exposure were studied using a microphysiometer modified to measure dopamine release, and Ca++ uptake. Upon exposure to OPCs, glucose and oxygen consumption decrease, while lactate production increases. We then constructed a large mass conserving, mechanistic map of neural signaling pathways implicated in OPC neurotoxicity, including PKC activation, Ca++ dynamics, and MAPK cascades. A comprehensive model of cellular metabolism was constructed encompassing glycolysis, TCA cycle and glycogen storage. The pathways were coupled to one another through shared species. Model predictions agree well with the metabolic and signaling experimental measurements over a range of OPC doses. Primary biomarkers and candidate control points were determined using a sensitivity analysis of the pathway model. The results indicate that PKC acting in both the RAS and MAPK pathway exerts the maximal control over the toxic response, making PKC an excellent initial candidate control point. Confirmatory experiments involving looking at gene expression across multiple cell lines is planned for the near future.

Dynamic Flux Balance Analysis Methods for Toxin Discrimination

In an ongoing effort, CFDRC has assembled a fully mass conserving, metabolic pathway model encompassing glycolysis, pentose phosphate, TCA cycle/oxidative phosphorylation, and calcium regulation of TCA dehydrogenases. The existing dynamic metabolic pathway model has been validated against metabolic measurements of Chinese hamster ovary (CHO) cells exposed to sodium fluoride, using the virtual microphysiometer simulation environment. The fluoride ion is a potent inhibitor of enolase, decreasing glycolytic flux and the production of pyruvate, ultimately causing the cell to increase the carbon flux through the TCA cycle in order to maintain ATP concentrations. The quantitative, dynamic pathway model accurately reproduces the dynamics of cellular response to fluoride ion exposure. The figure illustrates an application of the dynamic flux balance analysis (dFBA) algorithm in the discrimination of cellular response to toxins using MMP data. Botulinum toxin and fluoride cause a similar metabolic response (decrease in glucose and oxygen consumption, along with decreased lactate production), however a dFBA analysis reveals that cells exposed to fluoride respond by increasing the carbon flux to the TCA cycle relative to lactate, while cells exposed to botulinum toxin respond by increasing the lactate flux relative to TCA. Fluoride inhibits enolase and forces the cell to utilize the decreased carbon flux more efficiently. While the exact mechanism of botulinum toxin is unknown, this analysis indicates that the toxin decreases the carbon flux to the TCA cycle relative to lactate production.

Determination Of Critical Network Interactions: An Augmented Boolean Pseudo-Dynamics Approach

BNDTI algorithm used to generate a minimal pathway model of cellular response to neurotoxins. 272 species pathway model was generated using microarray data, and reduced using BNDTI. Green nodes in the large pathway model are the entry nodes, the blue node is Ca2+, and the red nodes are the output nodes. Reduced pathway interactions are colored by CEF value, with red being high values, and black being low.

Network theory has established that highly connected notes in regulatory networks (hubs), show a strong correlation with criticality in network function. While topological analysis is fully capable of identifying network hubs, it does not provide an objective method for ranking the importance of a particular node by relating its contribution to the overall cellular response. Towards this end, we have developed an augmented Boolean pseudo-dynamics approach to a priori determine the critical network interactions. The approach relies on network topology and the dynamic state information to determine the set of active pathways. The active pathways are used in conjunction with the key cellular properties of efficiency and robustness to rank the network interactions based on their importance in the sustenance of network function. In order to demonstrate the utility of the approach, we consider the well characterized guard cell signaling network in plant cells. An integrated analysis of the network revealed the critical mechanisms resulting in stomatal closure in the presence and absence of abscisic acid, in excellent agreement with published results.

3D initial conditions and (d) final folded coat. Colors in (b) and (d) represent the strain magnitude with high strain represent as red and low strain as blue.