Effects of higher-than-expected control population allele frequency on classification of loss-of-function variants in cancer susceptibility genes.

Abstract

A query was sent to the cancer predisposition gene variant database Cancer Variant Interpretation Group UK, on the nonsense variant in NM_032043.3(BRIP1):c.2392C>T,p.(Arg798Ter). The submitter classified this as a variant of uncertain significance, providing very strong variant effect evidence with the intention of adding supporting pedigree information, according to the guidelines used for classification. However, the relatively high population frequency in the UKB cohort of 367/439 920 (0.083%) was a concern as it is higher than expected for the disease frequency, which would reduce the predicted pathogenicity score. This situation highlights the increasing concerns over the use of population data in pathogenicity classification of truncating/loss-of-function (LoF) variants in known cancer predisposition genes, particularly since the addition of UKB control data. Here, we have conducted a series of case-control comparisons for common truncating variants in known breast/ovarian cancer-associated genes, as well as LZTR1-related schwannomatosis, to address this issue using our Manchester cancer screening population compared with controls in UKB data.Our data show strong ORs for these common truncating variants. We propose that for truncating variants in cancer susceptibility genes with a significant case-control OR, apparently conflicting population frequency evidence criteria should be avoided.