Stephen E. Stein National Institute of Standards and Technology (NIST) Office: 301-975-2505 FAX: E-mail: steve.stein@nist.gov Job Title: NIST Fellow Ph.D. in Physical Chemistry from the University of Washington, Seattle

Speaker: Stephen E. Stein, National Institute of Standards and Technology, 100 Bureau Drive, Gaithersburg, MD 20899

Topic: Building and Using Libraries of Peptide Mass Spectra

Place: Building 549, Auditorium, NCI at Frederick, Frederick, MD

Time: Tuesday, August 9, 2005, at 2:00 PM

Abstract: Current methods for identifying peptides from their MS/MS spectra employ sequence information only, ignoring the remainder of the the information contained in the full spectrum. An identification based on a full spectrum-to-spectrum match, used routinely in other areas of mass spectrometry, should identify peptides at a significantly higher level of confidence. This approach requires the availability of reference libraries of peptide mass spectra. This presentation will describe recently developed methods for creating such reference libraries from spectra that identify peptides identifications in LC-MS/MS experiments. In the first step, all spectra assigned to a single peptide ion by sequence search engines are combined to form a 'consensus spectrum'. This step removes noise peaks, derives measures of peak variability and assigns overall measures of spectrum reproducibility. The next step examines the consistency of each consensus spectrum with its assigned peptide sequence using a variety of independent relationships between sequence and fragmentation. Some of these factors are fraction of unassigned peaks, similarity of consensus and theoretical spectrum and correlations between y and b ions. These factors, along with original sequence matching scores, serve to generate a probability that the consensus spectrum is correct. After adding other global information, each spectrum is annotated and then combined to form a searchable reference library. In this way, a library of over 80,000 different peptide ions has been generated from publicly available and contributed data. Spectrum matching algorithms extended from algorithms routinely used to search libraries of electron ionization libraries will also be presented. By combining non-spectral information associated with source spectra, these libraries also offer a convenient means of re-using information from earlier experiments where the same peptide was identified.