Integrating epigenetic prior in dynamic Bayesian network for gene regulatory network inference

Abstract

Gene regulatory network (GRN) inference from high throughput biological data has drawn a lot of research interest in the last decade. However, due to the complexity of gene regulation and lack of sufficient data, GRN inference still has much space to improve. One way to improve the inference of GRN is by developing methods to accurately combine various types of data. Here we apply dynamic Bayesian network (DBN) to infer GRN from time-series gene expression data where the Bayesian prior is derived from epigenetic data of histone modifications. We propose several kinds of prior from histone modification data, and use both real and synthetic data to compare their performance. Parameters of prior integration are also studied to achieve better results. Experiments on gene expression data of yeast cell cycle show that our methods increase the accuracy of GRN inference significantly.