Despite recent advances in the application of data-dependent liquid chromatography/tandem mass spectrometry (LC/MS/MS) to the identification of drug metabolites in complex biological matrixes, a prior knowledge of the likely routes of biotransformation of the therapeutic agent of interest greatly facilitates the detection and structural characterization of its metabolites. Thus, prediction of the [M + H]+ m/z values of expected metabolites allows for the construction of user-defined MS(n) protocols that frequently reveal the presence of minor drug metabolites, even in the presence of a vast excess of coeluting endogenous constituents. However, this approach suffers from inherent user bias, as a result of which additional "survey scans" (e.g., precursor ion and constant neutral loss scans) are required to ensure detection of as many drug-related components in the sample as possible. In the present study, a novel approach to this problem has been evaluated, in which knowledge-based predictions of metabolic pathways are first derived from a commercial database, the output from which is used to formulate a list-dependent LC/MS(n) data acquisition protocol. Using indinavir as a model drug, a substructure similarity search on the MDL metabolism database with a similarity index of 60% yielded 188 "hits", pointing to the possible operation of two hydrolytic, two N-dealkylation, three N-glucuronidation, one N-methylation, and several aromatic and aliphatic oxidation pathways. Integration of this information with data-dependent LC/MS(n) analysis using an ion trap mass spectrometer led to the identification of 18 metabolites of indinavir following incubation of the drug with human hepatic postmitochondrial preparations. This result was accomplished with only a single LC/MS(n) run, representing significant savings in instrument use and operator time, and afforded an accurate view of the complex in vitro metabolic profile of this drug.