Liquid chromatography coupled with mass spectrometry (LC-MS/MS) is increasingly used for proteomic profiling of biofuids aiming at discovery of candidate biomarkers of disease. However the clinical utility of the technology is currently limited due to stochastic variation, restricted dynamic range of the instruments, and difficulties with the identification of peptide sequences underlying LC-MS peaks.  The difficulties can be partially alleviated by developing specialized statistical methods that make an efficient use of spectral information.

This talk will discuss two types of such methods for analysis of LC-MS/MS profiles. First, we will characterize the statistical properties of two algorithms, Peptide Prophet and decoy database search, that are currently popular for peptide sequence
identification from LC-MS/MS spectra. Second, we will discuss two approaches that combine quantitative LC-MS feature profiles and available peptide sequence identities to infer changes in abundance at the gene or a protein level. The approaches involve a random-effects model popular in meta-analysis (DerSimonian and Laird, 1986), and an
Empirical Bayes hierarchical model that provides a more realistic representation of the data at the expense of a greater computational effort. The methods are illustrated using computer simulations, and using a clinical study of cardiovascular disease where the models are judged by their ability to uncover previously known protein biomarkers.

Brief Bio:
After completing a bachelor's and master's degrees in Econometrics and Statistics at University of Geneva, Switzerland, Olga worked as a biostatistician at the University Hospitals of Geneva where she became interested in development and application of quantitative methods in molecular biology. She moved to Purdue and obtained a masters and PhD in Statistics under co-direction of Chris Bailey-Kellogg and Bruce Craig. As a graduate student she closely collaborated with biologists, chemists and computer scientists through projects in genomics and structural biology, statistical
consulting and an internship in a proteomics lab at Eli Lilly and Company. After graduation she took a position of post-doctoral associate in Ruedi Aebersold's lab at the Institute for Systems Biology in Seattle and worked on projects in computational
proteomics. In Fall 2006 she came back to Purdue as an Assistant Professor in Statistics and Computer Science.