A methodology is described for screening fragment ion spectra of peptides prior to database searching for protein identification. A software routine written in the Perl programming language was used to analyze data from previous Sequest database searches and develop a set of statistical descriptors that could be used to identify spectra not likely to yield useful results in a database search. A second Perl program used an evolutionary algorithm to optimize the criteria for each statistical descriptor and generate a formula for determining spectral quality. This formula was used by a third Perl program to screen data sets from four independent liquid chromatography tandem mass spectrometry runs. On the average, use of the screening program reduced the time required for a database search by 1/2 with little loss of useful information from the database search results.