A general framework for estimating the relative pathogenicity of human genetic variants

Abstract

Our capacity to sequence human genomes has exceeded our ability to interpret genetic variation. Current genomic annotations tend to exploit a single information type (e.g. conservation) and/or are restricted in scope (e.g. to missense changes). Here, we describe Combined Annotation Dependent Depletion (CADD), a framework that objectively integrates many diverse annotations into a single, quantitative score. We implement CADD as a support vector machine trained to differentiate 14.7 million high-frequency human derived alleles from 14.7 million simulated variants. We pre-compute “C-scores” for all 8.6 billion possible human single nucleotide variants and enable scoring of short insertions/deletions. C-scores correlate with allelic diversity, annotations of functionality, pathogenicity, disease severity, experimentally measured regulatory effects, and complex trait associations, and highly rank known pathogenic variants within individual genomes. The ability of CADD to prioritize functional, deleterious, and pathogenic variants across many functional categories, effect sizes and genetic architectures is unmatched by any current annotation.

Genomic analysis identifies masqueraders of full‐term cerebral palsy

“…In the autosomal recessive and X‐linked models, autosomal variants with homozygotes found in the databases (or variants on the Chromosome X with hemizygotes in the databases) were filtered. Sorting intolerant from tolerant (SIFT),20 (polymorphism phenotypinb v2) PolyPhen2 HVAR,21 the Genomic Evolutionary Rate Profiling (GERP),22 and combined annotation‐dependent depletion (CADD)23 scores were used to predict nucleotide‐level conservation and the impact of amino acid substitutions. After extracting rare variants (listed in Table S1), only the rare variants that were present in Online Mendelian Inheritance in Man (OMIM) linked to neurodevelopmental diseases were prioritized as candidate variants.…”

Whole genome sequencing of Caribbean Hispanic families with late‐onset Alzheimer's disease

“…Functional annotations were also used as filtering strategy, that is, variants were analyzed in order of increasing priority: exonic, inter‐genic, intronic and others (non‐coding RNA, 3′/5′ UTR, upstream/downstream gene). We also included any intronic or intergenic variants with a Combined Annotation Dependent Depletion (CADD)34 score of greater than 30 to include highly conserved and putative non‐exonic regulatory variants in the analyses. Further criteria for variant prioritization included, additional LOAD patients from different families carrying the variant and that the variant was either not present or present in very low frequency in the non‐demented family members.…”

Missense variants in TMEM67 in a patient with Joubert syndrome

“…In silico analyses to predict the effect of this missense change on protein structure and function all predict that this variant is likely to be tolerated; however, the predictions have not been confirmed by published functional studies. The CADD score for this variant is 2.2 15. This variant classifies as a VUS with ACMG PM2 and PP4 criteria being met 13…”

“…This variant has been reported in population databases with a frequency in ExAC of 1 of 120 000, which is consistent with an autosomal recessive condition (ACMG PM2). In silico analyses carried out to predict the effect of this missense change on protein structure and function suggest pathogenicity (SIFT: “deleterious”; PolyPhen‐2: “Probably damaging”; CADD score for this variant is 27.2)15 and this site is entirely conserved evolutionarily (ACMG PP3). As we have shown above, the patient's family history and clinical phenotype are entirely consistent with the proposed disease (ACMG PP4).…”