Fig. 1

The workflow of Bayesian network pathway construction. The procedure started with preprocessing of the gene expression data, including removal of non-expressed genes, protein sequence generation, gene ID conversion, gene data clustering, and knowledgebase generation. The protein sequence for all genes was then mapped to the KEGG database for initial pathway construction. The Bayesian network method was used to predict new metabolic pathways by integration of the gene-to-KO assignments, reference pathways, gene sampling, and knowledge constraints