Title: "Protein significance analysis in quantitative targeted and data-independent mass spectrometry-based"
Speaker: Olga Vitek, Department of Statistics and Department of Computer Science, Purdue University
Place: Rawls (RAWL) Hall 1086
Date: November 12, 2013; Tuesday
Time: 4:30pm
Abstract:
Mass spectrometry-based proteomics studies proteins in complex biological mixtures. Tremendous progress in experimental workflows and instruments now enables investigations that are quantitative, accurate and high-throughput. At the same time, the structure of stochastic variation and uncertainty presents methodological challenges of statistical nature.
This talk will briefly introduce modern targeted and data-independent mass spectrometry-based workflows. A statistical challenge in these workflows is the detection of differentially abundant proteins from spectral features with unequal information content. The unequal information manifests itself in the observed data in form of unequal technical variation. It can be an artifact of the experimental workflow, or it can be an intentional property of the experiment introduced to save the costs. We show that in these situations detailed statistical modeling using Empirical Bayesian approaches and regularization leads to accurate and interpretable solutions.