Proteomics aims to identify new markers and targets for the diagnosis and treatment of human disease. To realize this goal, methods and reagents are needed to profile proteins based on their functional properties, rather than mere abundance. Here, we describe a general strategy for synthesizing and evaluating structurally diverse libraries of activity-based proteomic probes. Quantitative screening of probe-proteome reactions coupled with bioinformatic analysis enabled the selection of a suite of probes that exhibit complementary protein reactivity profiles. This optimal probe set was applied to discover several enzyme activities differentially expressed in lean and obese (ob/ob) mice. Interestingly, one of these enzymes, hydroxypyruvate reductase, which was 6-fold upregulated in ob/ob livers, participates in the conversion of serine to glucose, suggesting that this unusual metabolic pathway may contribute to gluconeogenesis selectively in states of obesity.