Robust differential co-expression discovery: an insight into pharmacodynamics of tyrosine kinase inhibitor.

ABSTRACTStudies of gene expression in drug sensitive and resistant cell lines can help us understand mechanisms of drug resistance, side effects and uncover new potential drug targets. Such analyzes require methods that move beyond uni-variate tests of differential expression. Here we propose a new differential co-expression analysis framework to detect molecular mechanism of drug sensitivity. We started with the Fisher's Z transformation to measure the co-expression difference. Then we developed a novel method to conquer the common problems Fisher's Z transformation cannot handle -- heterogeneity of cell line tissue types, outlying samples or clusters and relatively small number of samples. We generated a tightly inter-linked network of dasatinib related genes, connected by differential co-expression and significant inter-correlation. We observed three main types of differential co-expression. We did literature survey of result gene pairs and confirmed that genes in the detected pairs are related to dasatinib, or related to each other. Such discoveries bring new insights on this second generation tyrosine kinase inhibitor.