Computational prediction of the pathogenic status of cancer-specific somatic variants
Loading...
Date
2019-08-20
Authors
Feizi, Nikta
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Background: In-silico classification of the pathogenic status of the ever-increasing somatic variants, identified through cancer genomic sequencing, is shown to be promising in promoting the clinical utilization of genetic tests. Unlike the majority of the available variant classification tools, the significance of the somatic variants in cancer initiation and progression demands the development of specified classifiers specialized for classifying cancer somatic variants.
Methods: We designed a support vector machine classifier trained exclusively based on the characteristics of cancer somatic single nucleotide variants. To label the pathogenic and non-pathogenic variants in the gold standard dataset, we innovated a bi-dimensional recurrence score to extract the pathogenic variants from the COSMIC dataset in a highly confident manner. Also, we collected non-pathogenic cancer somatic mutations mutually form 1000 genome project and COSMIC dataset.
Findings: In the task of classifying cancer somatic variants from both coding and non-coding regions of the genome our models outperformed the available state-of-the-art genomic variant classification tools.
Interpretation: Applying our models to genomic data from two major breast cancer cohort studies (METABRIC and TCGA-BRCA), we were able to identify a pathogenic somatic variant specifically potential for the prognosis of early onset of breast cancer in young women. We also demonstrated that the mutation status of MUC16 can be a potential prognostic factor for early onset of breast cancer.
Description
Keywords
Cancer somatic variants, computational classification, breast cancer, MUC16, TP53