Proteogenomics of Malignant Melanoma Cell Lines: The Effect of Stringency of Exome Data Filtering on Variant Peptide Identification in Shotgun Proteomics

The
identification of genetically encoded variants at the proteome
level is an important problem in cancer proteogenomics. The generation
of customized protein databases from DNA or RNA sequencing data is
a crucial stage of the identification workflow. Genomic data filtering
applied at this stage may significantly modify variant search results,
yet its effect is generally left out of the scope of proteogenomic
studies. In this work, we focused on this impact using data of exome
sequencing and LC–MS/MS analyses of six replicates for eight
melanoma cell lines processed by a proteogenomics workflow. The main
objectives were identifying variant peptides and revealing the role
of the genomic data filtering in the variant identification. A series
of six confidence thresholds for single nucleotide polymorphisms and
indels from the exome data were applied to generate customized sequence
databases of different stringency. In the searches against unfiltered
databases, between 100 and 160 variant peptides were identified for
each of the cell lines using X!Tandem and MS-GF+ search engines. The
recovery rate for variant peptides was ∼1%, which is approximately
three times lower than that of the wild-type peptides. Using unfiltered
genomic databases for variant searches resulted in higher sensitivity
and selectivity of the proteogenomic workflow and positively affected
the ability to distinguish the cell lines based on variant peptide
signatures.