Deleterious- and disease-allele prevalence in healthy individuals: insights from current predictions, mutation databases, and population-scale resequencing

Abstract

We have assessed the numbers of potentially deleterious variants in the genomes of apparently healthy humans by using (1) low-coverage whole-genome sequence data from 179 individuals in the 1000 Genomes Pilot Project and (2) current predictions and databases of deleterious variants. Each individual carried 281-515 missense substitutions, 40-85 of which were homozygous, predicted to be highly damaging. They also carried 40-110 variants classified by the Human Gene Mutation Database (HGMD) as disease-causing mutations (DMs), 3-24 variants in the homozygous state, and many polymorphisms putatively associated with disease. Whereas many of these DMs are likely to represent disease-allele-annotation errors, between 0 and 8 DMs (0-1 homozygous) per individual are predicted to be highly damaging, and some of them provide information of medical relevance. These analyses emphasize the need for improved annotation of disease alleles both in mutation databases and in the primary literature; some HGMD mutation data have been recategorized on the basis of the present findings, an iterative process that is both necessary and ongoing. Our estimates of deleterious-allele numbers are likely to be subject to both overcounting and undercounting. However, our current best mean estimates of ~400 damaging variants and ~2 bona fide disease mutations per individual are likely to increase rather than decrease as sequencing studies ascertain rare variants more effectively and as additional disease alleles are discovered.

Figure 1

Condel-Score Distribution for HGMD-Only DMs, Variants Only in the 1000 Genomes Low-Coverage Pilot, and Overlap Variants Shown are Condel scores for HGMD-only DMs (blue), variants only in the 1000 Genomes Low-Coverage Pilot data (yellow), and overlap variants (green). Condel scores range from 0 (the amino acid change is predicted to not damage the protein) to 1 (the amino acid change is predicted to damage the protein). The top decile of Condel scores is further subdivided in the panel on the right-hand side.

Figure 2

Representation of HGMD Variant Classes in the 1000 Genomes Low-Coverage Pilot (A) HGMD variants subdivided by functional category. The following abbreviations are used: DM, disease-causing mutation; FPs, in vitro or in vivo functional polymorphism; DFP, disease-associated polymorphism with additional supporting functional evidence; and DP, disease-associated polymorphism. (B) HGMD variants subdivided by variant type. The first four types are all SNPs.

Figure 3

Derived Allele-Frequency Spectra of Nonsynonymous and Synonymous Variants and HGMD SNPs in Three 1000 Genomes Low-Coverage Pilot Samples The following abbreviations are used: NSV, nonsynonymous variant; SV, synonymous variant; DM, disease-causing mutation; FP, in vitro or in vivo functional polymorphism; DFP, disease-associated polymorphism with additional supporting functional evidence; and DP, disease-associated polymorphism.