Statistical Analysis of Spectral Count Data Generated by Label-Free Tandem Mass Spectrometry-Based Proteomics

Label-free strategies for quantitative proteomics provide a versatile and economical alternative to labeling-based proteomics strategies. We have shown for different types of biological samples that spectral counting-based label-free quantitation is a promising avenue for biomarker discovery. Analyzing spectral count data generated from these studies is, however, not straightforward, as commonly used techniques for genomics data analysis are not suitable. In this book chapter, we describe three methods to analyze spectral count data, namely, cluster analysis, significance analysis of independent samples, and significance analysis of paired samples. For cluster analysis, we devise a novel distance measure between samples based on the Jeffrey divergence. This measure prevents highly abundant proteins from dominating others in contribution to the total sample difference. We employ the beta-binomial distribution for significance analysis of independent samples, which integrates both within-sample variation and between-sample variation into a single statistical model. Finally, the Mantel–Haenszel test is used for significance analysis of paired samples. We provide detailed illustrations of the steps involved in the analyses.