A functional genomic study on NCI's anticancer drug screen.

Pharmacogenomics requires massive computer exploration on heterogeneous databases. COMPARE, the gateway to the NCI's anticancer drug screen database, allows users to correlate drug-sensitivity profiles with a functional genomic database. However, most drugs of known molecular mechanism turn out to be uncorrelated with their molecular-target gene expression. Based on a novel statistical concept, liquid association, we develop an on-line system to identify candidate genes that intervene, confound and weaken the drug-gene correlation. The system takes queries and returns button-clickable tables of functionally associated genes for rerouting to knowledgebases such as Locus Link, OMIM and PubMed. We report results that link methotrexate resistance to DNA component biosynthesis, and taxol sensitivity to genes associated with human immunodeficiency virus infection. The drug-sensitivity database can be synergistically coanalyzed with gene expression data to study proteins of poorly understood physiological roles. When applied to the human prion, a cellular context embroidered with the gene expression network of Alzheimer disease is revealed.