Combining protein ratio p-values as a pragmatic approach to the analysis of multirun iTRAQ experiments.

iTRAQ labeling of peptides is widely used for quantitative comparison of biological samples using mass spectrometry. However, iTRAQ determined protein ratios have varying credibility depending on the number and quality of the peptide ratios used to generate them, and accounting for this becomes problematic particularly in the multirun scenario needed for larger scale biological studies. One approach to this problem relies on the use of sophisticated statistical global models using peptide ratios rather than working directly with the protein ratios, but these yield complex models whose solution relies on computational approaches such as stage-wise regression, which are nontrivial to run and verify. Here we evaluate an alternative pragmatic approach to finding differentially expressed proteins based on combining protein ratio p-values across experiments in a fashion similar to running a meta-analysis across different iTRAQ runs. Our approach uses the well-established Stouffer's Z-transform for combining p-values, alongside a ratio trend consistency measure, which we introduce. We evaluate this method with data from two iTRAQ experiments using plant and animal models. We show that in the specific context of iTRAQ data analysis this method has advantages of simplicity, high tolerance of run variability, low false discovery rate, and emphasis on proteins identified with high confidence.