An extended set of yeast-based functional assays accurately identifies human disease mutations

Abstract

We can now routinely identify coding variants within individual human genomes. A pressing challenge is to determine which variants disrupt the function of disease-associated genes. Both experimental and computational methods exist to predict pathogenicity of human genetic variation. However, a systematic performance comparison between them has been lacking. Therefore, we developed and exploited a panel of 26 yeast-based functional complementation assays to measure the impact of 179 variants (101 disease- and 78 non-disease-associated variants) from 22 human disease genes. Using the resulting reference standard, we show that experimental functional assays in a 1-billion-year diverged model organism can identify pathogenic alleles with significantly higher precision and specificity than current computational methods.

Figure 1.

Overview of the surrogate genetics platform for functional assessment of genetic variants. (A) Humanized yeast as a surrogate genetic system to evaluate functional effects of human genetic variation. (1) Disruption of a yeast gene gives rise to a yeast phenotype (e.g., decreased fitness). (2) The yeast phenotype is rescued by wild-type human alleles. (3) Functional effects of human variants are evaluated based on their ability to rescue the phenotype relative to the wild-type allele. A human variant may be deemed pathogenic if it cannot rescue the phenotype as well as wild type, or in special cases, where it exhibits better-than-wild-type rescue of the phenotype. (B) Overview of functional assessment of variants in disease-associated genes.

Figure 2.

Assessment of functional effect of missense variants. (A,B) Growth assays on solid media (A) and liquid media (B) for human tyrosyl-tRNA synthetase (YARS) gene variants. G41R and E196K are disease-associated variants; M431L and A339T are non-disease-associated variants. The yeast cells were temperature-sensitive mutants of the yeast TYS1 gene, expressing either wild-type or variant alleles of the YARS gene, or the GFP gene as a control. For solid growth assays, fivefold serial dilutions of yeast cells were spotted onto plates and incubated at 24°C and 36°C for 2 d. For liquid growth assays, approximately 10⁵ cells were inoculated in 100 μL liquid medium and the absorbance at 600 nm was read every 15 min. (C) Distribution of solid growth (FCS) scores for disease- and non-disease-associated variants. (D) Distribution of liquid growth (FCT) scores for disease- and non-disease-associated variants. (E) Precision and recall analysis for FCS and PolyPhen2 scores. (F) Overlap of disease-associated variants predicted to be deleterious by functional assays (orange) and PolyPhen2 (green). (G) Overlap of non-disease-associated variants predicted to be deleterious by functional assays (orange) and PolyPhen2 (green). As described in the text, analyses in this figure exclude FCT assays of PKLR variants and all assays of UROS variants.

Figure 3.

Exploring differences between functional assays and current pathogenicity annotation. (A–C) Among variants currently annotated as disease-associated, those classified as neutral by the FCS assay (apparent false negatives) overlap significantly with (A) variants that are also classified as neutral by PolyPhen2 (P = 1.6 × 10⁻⁵, Fisher's exact test); (B) disease-associated variants with unknown causality (P = 0.0032, Fisher's exact test); and (C) variants located at nonconserved residue positions. Enrichment in each case was relative to the corresponding frequency among variants called deleterious by the FCS assay (P = 0.036, Fisher's exact test). (D,E) Among non-disease-associated variants classified as deleterious by functional assays (apparent false positives), there were two variants for which structural context supported the finding of these variants to be deleterious: (D) DHFR[E62Q] and (E) UBE2I[F58C]. (Pink) Residues where variants occur; (cyan) functional sites.