Title: "False discovery rates in database-based identification of tandem mass spectra"
Speaker: Dr. Olga Vitek, Department of Statistics, Department of Computer Science, Purdue University
Place: Mechanical Engineering (ME) 161; January 15, 2008, Tuesday, 4:30pm

Abstract

Database-based identification of tandem mass spectra is now routinely used to study components of complex biological mixtures. However reliable identification of peptide and protein sequences from the spectra remains a challenge. The large amounts of spectra (tens or even hundreds of thousands in a single experiment), and noise and stochastic variation in the experimental procedure, frequently result in unknown and potentially high number of incorrect identifications.

A desirable output of the procedure is a list of identified spectra with a controlled expected proportion of incorrect identifications, i.e. False Discovery Rate (FDR). The talk will review the approaches that are currently used for estimation of False Discovery Rate, in particular target-decoy database search and the two-group model approach. We will discuss the properties of these approaches from the multiple model selection viewpoint, and show that the model selection framework can suggest alternative solutions to estimating FDR. We will illustrate the discussion using simulated and experimental datasets.


Associated reading:
Alexey I. Nesvizhskii, Olga Vitek and Ruedi Aebersold. 2007. Analysis and validation of proteomic data generated by tandem mass spectrom etry. Nature Methods. 4(10):787-797.
Click here for a full schedule of BIOINFORMATICS SEMINARS, past and present.