Abstract
BackgroundPerturbed posttranslational modification (PTM) landscapes commonly cause pathological phenotypes. The Cancer Genome Atlas (TCGA) project profiles thousands of tumors allowing the identification of spontaneous cancer-driving mutations, while Uniprot and dbSNP manage genetic disease-associated variants in the human population. PhosphoSitePlus (PSP) is the most comprehensive resource for studying experimentally observed PTM sites and the only repository with daily updates on functional annotations for many of these sites. To elucidate altered PTM landscapes on a large scale, we integrated disease-associated mutations from TCGA, Uniprot, and dbSNP with PTM sites from PhosphoSitePlus. We characterized each dataset individually, compared somatic with germline mutations, and analyzed PTM sites intersecting directly with disease variants. To assess the impact of mutations in the flanking regions of phosphosites, we developed DeltaScansite, a pipeline that compares Scansite predictions on wild type versus mutated sequences. Disease mutations are also visualized in PhosphoSitePlus.ResultsCharacterization of somatic variants revealed oncoprotein-like mutation profiles of U2AF1, PGM5, and several other proteins, showing alteration patterns similar to germline mutations. The union of all datasets uncovered previously unknown losses and gains of PTM events in diseases unevenly distributed across different PTM types. Focusing on phosphorylation, our DeltaScansite workflow predicted perturbed signaling networks consistent with calculations by the machine learning method MIMP.ConclusionsWe discovered oncoprotein-like profiles in TCGA and mutations that presumably modify protein function by impacting PTM sites directly or by rewiring upstream regulation. The resulting datasets are enriched with functional annotations from PhosphoSitePlus and present a unique resource for potential biomarkers or disease drivers.
Highlights
Perturbed posttranslational modification (PTM) landscapes commonly cause pathological phenotypes
Filtering the dbSNP dataset for disease-associated human variants, which result in missense alterations at the protein level, yielded a set of 18, 511 non-redundant mutations on 2532 proteins linked with more than 3000 different diseases
Splicing factor U2AF1 showed a recurrent mutation (S34F) in leukemia and lung adenocarcinoma resulting in the 7th highest hotspot score in our analysis
Summary
Integrating PTM sites with disease variants We retrieved PTM sites from PhosphoSitePlus (https:// www.phosphosite.org), somatic mutations from thousands of TCGA tumors from cBioPortal, and disease-associated SNPs from Uniprot and dbSNP (Materials and Methods). 138 (93%) agreed in sign (Fig. 6a and Additional file 8 A) Both Delta-Scansite and MIMP predicted a loss phosphorylation of T284 on p53 by Aurora B due to the alteration R282W found in 30 tumors across multiple cancer types. In addition to mutations leading to direct loss of regulating PTM sites of beta-catenin as described above, both approaches predicted reduced phosphorylation on S33 by GSK, triggered by three hotspot mutations on a flanking residue that is itself a phosphosite (S37C, S37F, and S37A). These mutations were found in primarily endometrial cancer tumors.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have