Defining the Players in Higher-Order Networks: Predictive Modeling for Reverse Engineering Functional Influence Networks

Abstract

Determining biological network dependencies that can help predict the behavior of a system given prior observations from high-throughput data is a very valuable but difficult task, especially in the light of the ever-increasing volume of experimental data. Such an endeavor can be greatly enhanced by considering regulatory influences on co-expressed groups of genes representing functional modules, thus constraining the number of parameters in the system. This allows development of network models that are predictive of system dynamics. We first develop a predictive network model of the transcriptomics of whole blood from a mouse model of neuroprotection in ischemic stroke, and show that it can accurately predict system behavior under novel conditions. We then use a network topology approach to expand the set of regulators considered and show that addition of topological bottlenecks improves the performance of the predictive model. Finally, we explore how improvements in definition of functional modules may be achieved through an integration of inferred network relationships and functional relationships defined using Gene Ontology similarity. We show that appropriate integration of these two types of relationships can result in models with improved performance.

Modeling Dynamic Regulatory Processes in Stroke

“…Considering only the LPS preconditioning time course, the correlation was somewhat higher at 0.78, as were the other time courses alone: CPG-ODN (0.80), ischemic preconditioning (0.82), saline treatment (0.92), and sham surgery (0.81). We have previously reported similar results using models developed with the Inferelator that consider individual genes as regulatory influences, rather than entire clusters [23], [26]. These models can be used to perform limited types of simulations, such as predicting expression of target clusters after in silico deletion of a regulatory influence.…”

“…This algorithm uses an approach called L1 error regression (also known as Lasso) to choose a parsimonious set of regulatory influences that can explain the expression of each cluster maximally [21], [22]. We have previously described elements of this approach in this and other systems [23], [24], [25], [26]. The result is a model abstracted from high-throughput transcriptomic data that mathematically expresses the relationships between clusters.…”

“…Although the Inferelator is capable of incorporating time course data explicitly in its inference process, we chose to treat data from each time point as a steady-state measurement since the time steps were relatively long and of variable length. We assessed the performance of the model using a cross-validation approach in which multiple models are trained on subsets of the data, by leaving out the data corresponding to a treatment time course, and then evaluated on the excluded time course data as previously described [23], [24], [25], [26]. Prediction of the expression levels for a cluster in this steady-state model requires input of expression levels for the regulatory influences, in this case the clusters themselves, for each condition being predicted.…”

The Use of High-Throughput Technologies to Investigate Vascular Inflammation and Atherosclerosis

“…173 In addition, a predictive network model, evaluating the transcriptome of whole blood from a mouse model of neuroprotection in ischemic stroke, was described, and it proved to be able to precisely predict mammalian system behavior under novel conditions. 174 Far from being complete, these examples demonstrate how network modeling has contributed to the analysis of highthroughput data sets and the identification of new cardiovascular disease-associated genes and pathways that would not have been discovered with the traditional, single-analysis approaches. In this way, networks could trigger the development of new diagnostic markers (eg, of atherosclerosis) by revealing targets that accurately predict lesion vulnerability.…”

“…44 Accordingly, systems biology and predictive network modeling may be used to develop concrete clinical applications helping to improve, for example, patient selection or monitoring of stroke preventive intervention. 174 In addition, the subdivision of networks of multifactorial diseases in subnetworks or modules indicates that multiple genes and pathways probably need to be tackled together to treat the disease as efficient as possible, and could offer help in the design of a combination therapy. For more detailed information on network modeling, and an overview of currently available tools and databases, we refer to previous in-depth reviews.…”