A novel method to infer genetic regulatory networks from microarray experiments based on Sparse Neighbor Bayesian Network Learning

Though a number of advances have been made in the past decade to cure fatal diseases such as Cancer, we still lack a good therapeutic target for a vast majority of fatal diseases. The availability of large-scale high- throughput biological data has motivated a number of researchers to apply computational methods to systematically model the behavior of biological networks. These networks built from Big Biological data help in identifying novel therapeutic targets for the fatal diseases. Such networks provide a "snapshot" of the interactions and key biological features of the disease system and are therefore, crucial for untangling the molecular networks underlying a disease before we can discover any effective therapeutics.;This thesis aims at developing novel computational algorithm, called sparse neighbor Bayesian network learning, in systems biology to reverse-engineering the molecular interaction networks in-silico using Machine Learning and Bayesian Networks for various diseases. In the first part of this thesis, we developed a novel ensemble supervised learning framework to reveal the neighborhood of the genes in the network. In this second part, we integrate this neighbor network with the popular Bayesian network structure learning method to improve the accuracy of the standard Bayesian network learning. We verified our algorithm on standard benchmark networks from DREAM4 competition, we showed that our algorithm significantly outperform existing standard Bayesian network method in reconstructing the genetic regulatory networks.