Analyzing Phenotype Microarray Data for Escherichia coli Using an Infinite Relational Model

To elucidate the dependence of gene function on the environmental conditions, we focused on quantitative data from a phenotype microarray (PM) of the wild type and ca. 300 single-gene knockout mutants of Escherichia coli K-12 cultured under various medium conditions. We developed an infinite relational model (IRM) applicable to three-valued relational data and applied it to the PM data. The results of gene ontology (GO) analysis showed that mutants with deletion of the genes that encode the enzymes threonine synthase and methionine synthase exhibited reduced cell growth in medium containing amino acids as a nutrient source. By comparing the number and degree of overlap of clusters enriched in certain GO terms obtained by IRM and other biclustering methods, we confirmed the effectiveness of our IRM for omics data analysis.