Validation of Proteogenomic Annotations

In Proteogenomics pipelines, tandem mass spectra are searched against theoretical spectra of peptide sequences derived from genomic and transcriptomic data, to identify mutations, splice variants, and proteins disrupted by structural variation. Proteogenomics is a specific focus of cancer proteome analysis (including the CPTAC projects), as cancer is marked by somatic genome lesions. While many somatic mutations have been identified using cancer proteogenomic pipelines, the increased sensitivity comes at the expense of false-discoveries. Standard p-value and FDR controls are not effective as the decoy databases do not account for all mutations. Here we ask if “a proteogenomically discovered mutation has a simpler explanation as a reference (modified) peptide”. We describe a new method, Proteogenomics-Event-Validation (PEV) to address this. PEV takes as input a list of mutated peptide spectrum matches (PSMs). For each PSM, it enumerates reference (modified) isobaric peptides that match the spectrum. The reference peptides are prioritized and scored using a Bayesian framework to determine if the original identification was better than the reference based alternative.

We applied PEV to spectra from 95 CPTAC colorectal samples previously searched against a database constructed from 86 TCGA RNA-seq data. PEV analysis on the PSMs from the 1035 mutated peptides suggested 543 peptides with at least one reference (modified) alternative. PEV used MS-GF+ to rescore the mutated and alternative peptides. In a vast majority of the cases, the mutated peptide remained the best explanation suggesting robustness of proteogenomics discovery. For example, the T->S mutation in peptide K.VQLPTE(T->S):LQELLDLHR.D appears correct because the best alternative explanation required Oxidation of Proline. However, in a small number of cases, the alternative explanation dominated. For example, an “N->D” mutation was better explained by Deamidation of Asn. PEV promises to be a useful resource for validating proteogenomic discoveries.