Application of protein set enrichment analysis to correlation of protein functional sets with mass spectral features and multivariate proteomic tests

Mass spectral data from multiple samples are suitable for a hypothesis-free development of clinically useful multivariate tests using modern machine learning techniques. However, the transition from discovery to adoption of proteomic tests has proved challenging. Slow adoption of these tests in clinical practice is, in part, related to insufficient understanding of the biological mechanisms underlying multivariate tests developed based on correlative studies. While identification of individual proteins may provide important insights, elucidation of concerted relationships of sets of proteins with biological pathways can better reflect complex phenomena, such as cancerogenesis and response to treatment. Protein set enrichment analysis (PSEA) allows identification of associations of mass spectral features or test classifications with biological function by looking for consistent correlations across a group of proteins.

We evaluated the utility of PSEA for exploring the biological information content of mass spectra, using a sample set with mass spectral data and matched protein expression information. This made it possible to detect significant biological associations with mass spectral peaks without identifying their protein constituents. We demonstrated that the method produces reproducible associations and can be used for elucidation of the mechanisms of action associated with two previously developed multivariate mass spectrometry-based tests. Significant correlations with several host immune response-related processes were found on the level of individual mass spectral features and with test classifications. The results illustrate the utility of the PSEA approach applied to mass spectral data as a method for elucidating biological mechanisms underlying phenotypes related to different physiological states of the organism.