The structural characterization of proteins and peptides isolated in minute quantities requires the most efficient use of available sample. A mass spectrometer data system was programmed to continuously evaluate incoming liquid chromatography/mass spectrometry data against a user-defined array of information. The resulting conclusions were used to automatically set and modify acquisition parameters in real time to collect collision-induced dissociation spectra for selected ions (tandem mass spectrometry). This approach has provided a mechanism to target specific subsets of masses in a complex mixture and/or to discriminate selectively against masses that are known or not of interest. Masses of contaminants or peptide masses derived from known proteins can be automatically recorded and removed from further consideration for collision-induced dissociation analysis. Once recorded, these "libraries" of masses can be used across multiple analyses. This technique directs the mass spectrometer data system to focus on the analysis of masses significant to the user, even if their signal intensities are well below the intensities of contaminating masses. When combined with a database search program to correlate tandem mass spectra to known protein sequences, the identity of the protein can be established unequivocally by using less than 100 fmol of sample.