Abstract
It is an important issue to infer the causal relation among the genes of interested. In this study, we proposed a four-step strategy for inferring the causal relation. First, we selected closely related gene modules based on their correlation structure of gene expression. Secondly, we used the multivariate sparse partial least squares (M-SPLS) regression to identifying the corresponding SNPs underlying each gene module. The third step, inferred the causal relation of sub-networks with each gene module and its corresponding SNPs by Bayesian network. Finally, we assembled the sub-networks by a ranking strategy to reconstruct the pathway associated with cancer metastasis.