Testing the Performance of the MutPred2 Variant Pathogenicity Predictor on a Set of BRCA1 and BRCA2 Variants and the Relevance of Solvent Accessibility in Predicting Pathogenicity

Abstract

Mutations in the BRCA1 and BRCA2 genes are known to be
associated with an increased risk of breast and ovarian cancer.
There are variants of BRCA1 and BRCA2 genes that have been
clinically assessed for pathogenicity, yet there are new and
undiscovered variants that have not. These gene variants are
classified as Variants of Uncertain Significance (VUSs) and
individuals who have a variant in this class could possibly be in
risk of developing cancer. In absence of clinical data, In Silico
Prediction Algorithms are techniques that predict the probability
of a variant being pathogenic by using parameters such as
protein structure of the variant and amino acid substitutions. The
purpose of this study is to test the performance of the In Silico
prediction software known as MutPred2 and also determine if
there is an advantage in using solvent accessibility as a predictor
for variant pathogenicity. To address this question we applied the
MutPred2 prediction software on a set of single-nucleotide
change BRCA1 and BRCA2 variants that were added to the
BRCA Exchange after the MutPred2 release date. Performance
is based on how MutPred2 output compares to pathogenicity
scores in the ENIGMA and ClinVar database which has data on
clinical-based studies. We identified outliers that MutPred2 had
difficulty in predicting pathogenicity and check to see if solvent
accessibility can aid in predicting pathogenicity. If MutPred2 and
solvent accessibility predictions compare well to ENIGMA and
ClinVar classifications, then we have evidence that solvent
accessibility is valuable in predicting pathogenicity.

Presentation