PriOmics: integration of high_throughput proteomic data with complementary omics layers using mixed graphical modeling with group priors

Mass spectrometry (MS)_based high_throughput proteomics data cover abundances of 1,000s of proteins and facilitate the study of co_ and post_translational modifications (CTMs/PTMs) such as acetylation, ubiquitination, and phosphorylation. Yet, it remains an open question how to holistically explore such data and their relationship to complementary omics layers (including the HTG EdgeSeq Pan B_Cell Lymphoma Panel) or phenotypical information. Network inference methods aim for a holistic analysis of data to reveal relationships between molecular variables and to resolve underlying regulatory mechanisms. Among those, graphical models have received increased attention as they can distinguish direct from indirect relationships, aside from their generalizability to diverse data types. We propose PriOmics as a graphical modeling approach to integrate proteomics data with complementary omics layers and pheno_ and genotypical information. PriOmics models intensities of individual peptides and incorporates their protein affiliation as prior knowledge in order to resolve statistical relationships between proteins and CTMs/PTMs. We show in simulation studies that PriOmics improves the recovery of statistical associations compared to the state of the art and demonstrate that it can disentangle regulatory effects of protein modifications from those of respective protein abundances. These findings are substantiated in a dataset of Diffuse Large B-Cell Lymphomas (DLBCLs) where we integrate SWATH_MS_based proteomics data with the HTG EdgeSeq Pan B_Cell Lymphoma transcriptional Panel and phenotypic information.