Exploring penetrance of clinically relevant variants in over 800,000 humans from the Genome Aggregation Database

Abstract

Incomplete penetrance, or absence of disease phenotype in an individual with a disease-associated variant, is a major challenge in variant interpretation. Studying individuals with apparent incomplete penetrance can shed light on underlying drivers of altered phenotype penetrance. Here, we investigate clinically relevant variants from ClinVar in 807,162 individuals from the Genome Aggregation Database (gnomAD), demonstrating improved representation in gnomAD version 4. We then conduct a comprehensive case-by-case assessment of 734 predicted loss of function variants in 77 genes associated with severe, early-onset, highly penetrant haploinsufficient disease. Here, we identify explanations for the presumed lack of disease manifestation in 701 of 734 variants (95%). Individuals with unexplained lack of disease manifestation in this set of disorders are rare, underscoring the need and power of deep case-by-case assessment presented here to minimize false assignments of disease risk, particularly in unaffected individuals with higher rates of secondary properties that result in rescue.

Fig. 1. Representation of ClinVar variants in 807,162 individuals in gnomAD v4.

a Variant count of ClinVar variants in ClinVar (gray) vs. gnomAD (blue) in each classification category pathogenic/likely pathogenic (P/LP), variant of uncertain significance (VUS), benign/likely benign (B/LB) or with conflicting classifications. b Percentage of ClinVar variants reported in gnomAD in at least one individual. c The proportion of ClinVar variants in gnomAD by variant type within each clinical significance classification and, d by allele frequency bin. e Total number of P/LP variants within each allele count bin (including variants absent from gnomAD for comparison). f The inheritance pattern of the gene harboring the P/LP variants in gnomAD versus all variants in ClinVar, X-linked (XL), Autosomal dominant (AD), Autosomal recessive (AR).

Fig. 2. The p.Gly45Glu GJB2 variant associated with severe pediatric disease is rescued by a modifying variant in the East Asian genetic ancestry group.

a Schematic of the rescue mechanisms where the modifying p.Tyr136Ter nonsense variant in cis with the dominant-negative disease-causing p.Gly45Glu missense variant, results in incomplete penetrance by mono-allelic expression of the reference allele only (green). b Paired individual-level read data of the two variants occurring in cis. c East Asian (n = 34) and Admixed American (n = 1) individuals carrying both the pathogenic p.Gly45Glu variant (left panel) and the p.Tyr136Ter modifying variant (right panel).

Fig. 3. Deep assessment of 734 predicted high-confidence (HC) loss-of-function (pLoF) variants found in 77 haploinsufficient (HI) genes in 76,215 genomes.

a Schematic of how modifying variants (black) may result in lack of a phenotype. b Filtering approach. 1KG: 1000 Genomes Project, AB: Allele balance, CH: clonal hematopoesis, MANE: Matched Annotation from NCBI and EMBL-EBI. c Variant count per HI gene colored by (also for e-g): explained (blue), uncertain (gray), unexplained (red). The number indicates unexplained variants (red). d The explanation for evading LoF in 511 of 734 (69.6%) of variants, MNV: multi-nucleotide variant, pext: per-base-expression score, tx: transcript, cons: conservation, ‘last exon 25’ and ‘SpliceAI in frame exon 25’: less than 25% protein removed; ‘AB below 25’: allele balance <25%. e Comparison of outcome in different gene sets, this set (left), heterozygous pLoF variants in autosomal recessive (AR) disease genes in gnomAD v2 (middle), all homozygous (hom) pLoF variants in gnomAD v2 (right). f Allele frequency (AF) of variants in this set by LoF curation outcome (log₁₀ scale), lower AF for unexplained (n = 137) than explained (n = 511), p = 0.0378 two-sided Student’s t-test. g The proportion of pLoF variants explained, uncertain, or unexplained within each ClinVar clinical classification category pathogenic/likely pathogenic (P/LP), uncertain significance (VUS), benign/likely benign (B/LB).

Fig. 4. Splicing quantitative trait loci (sQTL) analysis.

a Schematic of alternative splicing as a mechanism for incomplete penetrance. b Labels for each of the nine predicted loss of function (pLoF) variants. Bars are colored by gene, labels indicate variant count per gene. Tx: transcript, pext: per base expression score, AB: allele balance. c Overview of pLoF variants in MEF2C in ClinVar and in gnomAD v4 (all variants, including exomes). Black box marks region of reduced mean pext score (from v2, blue bar, for one exon the score is not available [N/A]) and enrichment of MEF2C pLoF variants in gnomAD.