Phosphatases are crucially involved in cellular processes by dephosphorylating cellular components. We describe a structure-based classification scheme for all active human phosphatases that reveals previously unrecognized relationships between them. By collating protein and nonprotein substrates and integrating colocalization and coexpression data, we generated a human phosphatase-substrate network. Analysis of the protein sequences surrounding sites of dephosphorylation suggested that common recognition mechanisms may apply to both kinases and a subset of phosphatases. Analysis of three-dimensional substrate recognition by protein phosphatases revealed preferred domains in the substrates. We identified phosphatases with highly specific substrates and those with less specificity by examining the relationship between phosphatases, kinases, and their shared substrates and showed how this analysis can be used to generate testable hypotheses about phosphatase biological function. DEPOD (human DEPhOsphorylation Database, version 1.0, http://www.DEPOD.org) is an online resource with information about active human phosphatases, their substrates, and the pathways in which they function. The database includes links to kinases and chemical modulators of phosphatase activity and contains a sequence similarity search function for identifying related proteins in other species.