Improving current microbial pathway models by statistical modeling of phenotype array experiments

Hundreds of bacterial genomes have been sequenced, but only a fraction of the genes have known biochemical function. Advances in cellular phenotyping promise to narrow the gap and improve current annotations. Phenotype MicroArrays (PMs) simultaneously measure the response of an organism against thousands of conditions, and thus provide a high-throughput means to characterize microbial phenotypes and metabolism. The PM technology is completely automated, but current analysis methods involve time consuming visual inspection of the data, and thus present a bottleneck. We propose rigorous statistical methods to automatically assess the results of PM experiments, and to incorporate the functional information gained from PMs with existing knowledge from complementary genomic and proteomic platforms. The impact will be an improved data mining of high-throughput phenotype experiments, as well as an unprecedented ability to characterize microbes and improve current microbial pathway models.