Massive proteogenomic reanalysis of publicly available proteomic datasets of human tissues in search for protein recoding via adenosine-to-inosine RNA editing

Abstract

The proteogenomic search pipeline developed in this work has been applied for re-analysis of 40 publicly available shotgun proteomic datasets from various human tissues comprising more than 8,000 individual LC-MS/MS runs, of which 5442 .raw data files were processed in total. The scope of this re-analysis was focused on searching for ADAR-mediated RNA editing events, their clustering across samples of different origin, and classification. In total, 33 recoded protein sites were identified in 21 datasets. Of those, 18 sites were detected in at least two datasets representing the core human protein editome. In agreement with prior art works, neural and cancer tissues were found being enriched with recoded proteins. Quantitative analysis indicated that recoding of specific sites did not directly depend on the levels of ADAR enzymes or targeted proteins themselves, rather it was provided by differential and yet undescribed regulation of interaction of enzymes with mRNA. Nine recoding sites conservative between human and rodents were validated by targeted proteomics using stable isotope standards in murine brain cortex and cerebellum, and an additional one was validated in human cerebrospinal fluid. In addition to previous data of the same type from cancer proteomes, we provide a comprehensive catalog of recoding events caused by ADAR RNA editing in the human proteome.

Proteogenomics analysis of human tissues using pangenomes

“…We removed 6.8% of the identified PSMs using MS2PIP. These results are similar to a previous study by Levitsky et al 47 that removed ∼5% of PSMs using a similar approach with the prediction tool PROSIT 60 . Although SNR and M2PIP removed 21.8% of the low-quality PSMs, none of the algorithms validates the quality of a particular amino acid variant within a peptide identification.…”

“…We removed 6.8% of the identified PSMs using MS2PIP. These results are similar to a previous study by Levitsky et al 47 that removed ~5% of PSMs using a similar approach with the prediction tool PROSIT 60 .…”

“…We applied the MS2PIP algorithm independent of the search engine to keep peptides with a higher number of b and y ions annotated (PXD010154 - Figure 2C , PXD016999 - Supplementary Note 4 - Figure 8 ). Different from previous approaches that use b and y ions Pearson correlations to filter peptides 47 , we calculated a weighted score by combining the b and y ions scores and giving more weight to the y-ions. Using the weighted MS2PIP score a total of 6.9% and 6.2% of the GCA novel PSMs were removed ( Figure 2A ).…”

“…Peptide retention time directly depends on the amino acid sequence and thus it can be used as an additional confirmation of identifications. Multiple studies have used retention time prediction to filter false-positive identifications for peptides with significant differences between the experimental and predicted retention time 21, 47, 48 since the predicted retention from the peptide sequence should be closer to the experimental retention; if the difference is high, the identified peptide is more likely to be a false positive. DeepLC could be used with existing models from previous datasets or by using the experimental retention from the same dataset via transfer learning.…”

Multi-tissue transition of A-to-I RNA editing pattern and its regulatory relevance in transcription, splicing, and translation during development

“…Notably, significant up-regulation of two missense editing c.284A>G (p.K95R) and c.232A>G (p.R78G) in the insulin-like growth factor binding protein 7 (Igfbp7) gene were observed in the mouse brain, heart and kidney at the adult stage compared to the embryonic stage. These two conserved missense sites were also found in other mammals, such as humans and pigs 36 .…”