Title: "Statistical framework for protein quantification with Selected Reaction Monitoring (SRM) experiments"
Speaker: Veavi Chang, Department of Statistics, Purdue University
Place: HORT 117; March 8, 2011, Tuesday, 4:30pm

Abstract

Selected Reaction Monitoring (SRM) workflow enables sensitive and accurate protein quantification in complex biological mixtures. Modern SRM-based investigations involve replicate biological samples and increasingly complex experimental designs. They simultaneously profile large numbers of proteins, where each protein is represented by several peptide ions and several transitions. A variety of computational tools are now available to detect, store, quantify and visualize the transitions. However, there is currently no consensus on how to appropriately handle the repeated quantitative measurements on a protein in a sample, and how to derive protein-level conclusions across all labels, peptides, transitions, samples and conditions.

Deriving protein-level conclusions from SRM experiments is challenging, in large part, due to the stochastic variation and uncertainty in interpretation of the spectra. Statistical modeling allows us to derive objective an reproducible conclusions in such situations. Development of a probabilistic model is a key analysis step, because the choice of the model can strongly affect the conclusions. It is also one of the most challenging steps, as a single off-the-shelf model cannot accurately reflect all experimental designs and workflows.

We propose a general statistical modeling framework for protein quantification, based on linear mixed-effects models. It is applicable to most experimental designs, and to both label-free and label-based SRM workflows. We illustrate the utility of the framework in two investigations, which both aim at comparing protein abundances between conditions, but have different designs. We further illustrate the sensitivity and specificity of the framework using a spike-in experiment and a series of simulated datasets. We show that the proposed framework is sensitive and specific, can be adapted to the specifics of the experiments, and helps choosing experimental design.

Associated Reading:
P. Picotti, B. Bodenmiller, L.N. Müller, B. Domon and R. Aebersold. Full dynamic range proteome analysis of S. cerevisiae by targeted proteomics. (2009) Cell, 138(4):795-806, 2009.



Click here for a full schedule of BIOINFORMATICS SEMINARS, past and present.