Comparing gene expression similarity metrics for connectivity map

Connectivity map data and associated methodologies have become a valuable tool in understanding drug mechanism of action (MOA) and discovering new indications for drugs. The basic idea of connectivity map is to measure the similarity between disease gene expression signatures and compound-induced gene expression signatures. We evaluate different gene expression profile similarity metrics by comparing their ability to predict a compound's chemical grouping using the Anatomical Therapeutic Chemical (ATC) drug classification system. The results show that our simple eXtreme sum (XSum) and eXtreme cosine (XCos) measures perform significantly better than the standard Kolmogorov-Smirnov (KS) statistic in term of area under the Receiver Operating Characteristic (ROC) curve (AUC) and partial AUC.