Reconstructing gene regulatory network based on candidate auto selection method

Abstract

The reconstruction of gene regulatory network (GRN) is a great challenge in systems biology and bioinformatics, and methods based on Bayesian network (BN) draw most of attention because of its inherent probability characteristics. As NP-hard problems, most of the BN methods often adopt the heuristic search, but they are time-consuming for biological networks with a large number of nodes. To solve this problem, this paper presents a Candidate Auto Selection algorithm (CAS) based on mutual information and breakpoint detection to limit the search space in order to accelerate the learning process. The proposed algorithm automatically restricts the neighbors of each node to a small set of candidates before structure learning. Then based on CAS algorithm, we propose a globally optimal greedy search method (CAS+G), which focuses on finding the high-scoring network structure, and a local learning method (CAS+L), which focuses on faster learning the structure with small loss of quality. Results show that the proposed CAS algorithm can effectively identify the neighbor nodes of each node. In the experiments, the CAS+G method outperforms the state-of-the-art method on simulation data for inferring GRNs, and the CAS+L method is significantly faster than the state-of-the-art method with little loss of accuracy. Hence, the CAS based algorithms are more suitable for GRN inference.