Evaluation of predicted small-molecule targets.

For each experiment we evaluated the list of targets inferred by our model compared to a list of known targets, defined for k252a as human protein kinases [32] and for SB-202190 as those identified by Karaman et al.[33]. “Detected” refers to all proteins detected by LC-MS/MS and “significant” refers to those inferred as significant by our model (local fdr threshold .01). Precision was calculated as the percent of proteins inferred as significant by our model that were also known targets, and recall was calculated as the percent of known targets detected by LC-MS/MS that were also inferred as significant by our model. Because our lists of known targets were incomplete, precision statistics were likely underestimated. Targets with additional evidence of binding each small-molecule are annotated in Table S1.