CADD score has limited clinical validity for the identification of pathogenic variants in noncoding regions in a hereditary cancer panel

Abstract

PURPOSESeveral in silico tools have been shown to have reasonable research sensitivity and specificity for classifying sequence variants in coding regions. The recently-developed Combined Annotation Dependent Depletion (CADD) method generates predictive scores for single nucleotide variants (SNVs) in all areas of the genome, including non-coding regions. We sought to determine the clinical validity of non-coding variant CADD scores.METHODSWe evaluated 12,391 unique SNVs in 624 patient samples submitted for germline mutation testing in a cancer-related gene panel. We compared the distributions of CADD scores of rare SNVs, common SNVs in our patient population, and the null distribution of all possible SNVs stratifying by genomic region.RESULTSThe median CADD scores of intronic and nonsynonymous variants were significantly different between rare and common SNVs (p<0.0001). Despite these different distributions, no individual variants could be identified as plausibly causative among rare intronic variants with the highest scores. The ROC AUC for non-coding variants is modest, and the positive predictive value of CADD for intronic variants in panel testing was found to be 0.088.CONCLUSIONFocused in-silico scoring systems with much higher predictive value will be necessary for clinical genomic applications.

GAVIN: Gene-Aware Variant INterpretation for medical sequencing

“…There are several possible explanations for potential non-informativity of CADD scores. It may have bias towards the in silico tools and sources it was trained on, limiting their predictiveness for certain genomic regions or disease mechanisms [25]. Furthermore, calibration of pathogenic variants could be difficult in genes with high damage tolerance, i.e.…”

A pipeline combining multiple strategies for prioritizing heterozygous variants for the identification of candidate genes in exome datasets

“…Exomiser v2 only admits HPO term so to compare with Phevor; Disease Ontology Terms and Gene Ontology Terms were not used. No threshold value was applied in these analyses since the list of variants is generated from pre-filtered variants from VAAST.Combined Annotation-Dependent Depletion (CADD)CADD v1.3 [28, 59] is pre-computed score database that is based on classifier algorithms. The major goal of CADD is to predict the deleterious, functionally significant and pathogenic variants from diversified class of variants by integrative annotations.…”

GAVIN - Gene-Aware Variant INterpretation for medical sequencing

“…There are several possible explanations for potential non-informativity of CADD scores. It may have bias towards the in silico tools and sources it was trained on, limiting their predictiveness for certain genomic regions or disease mechanisms [25]. Fig.…”