EAPSI: Reconstructing gene regulatory networks using the neighborhood sampler

Tran Thanh V, Rochester NY

Investigators

Abstract

Genetic regulatory networks (GRNs) are crucial to understanding how genes, proteins and other DNA components interact to carry out cellular processes. Successful reconstruction of GRNs allows prediction of cellular behavior in response to external stimuli such as stress and carcinogens. Additionally, identification of influential genes in the network can illuminate the development of certain diseases and inform the design of targeted therapy. Therefore, good models that accurately reflect the workings of a GRN are important for health research. The aim of this project is to build GRNs that faithfully capture real network behavior. Collaboration with Drs. Sarah Boyd and Jonathan Keith- experts in the field of molecular biology and computational biology- at Monash University in Australia will provide for a new technique to build networks and data for testing the validity of the method. Although there are different ways to model GRNs, Boolean networks are simple to implement and provide a qualitative interpretation of genetic interactions. Because genetic data often contain noise and error, a Bayesian approach to network inference accommodates both data and model uncertainty. However, Bayesian inference is little developed in Boolean network reconstruction. The Neighborhood Sampler is a new Bayesian technique that uses Monte Carlo Markov Chain (MCMC) sampling in discrete model spaces. Because the sampler is efficient and easy to implement, it will be used to learn GRNs. The results of the model fit are estimates of the existence and type of regulatory relationships (up regulation or down regulation) in the network. The performance of the Boolean network model and the Neighborhood Sampler will be assessed with data and network simulations. This NSF EAPSI award is funded in collaboration with the Australian Academy of Science.