Reassessment of Mendelian gene pathogenicity using 7,855 cardiomyopathy cases and 60,706 reference samples

Abstract

Purpose: The accurate interpretation of variation in Mendelian disease genes has lagged behind data generation as sequencing has become increasingly accessible. Ongoing large sequencing efforts present huge interpretive challenges, but they also provide an invaluable opportunity to characterize the spectrum and importance of rare variation.

Methods: We analyzed sequence data from 7,855 clinical cardiomyopathy cases and 60,706 Exome Aggregation Consortium (ExAC) reference samples to obtain a better understanding of genetic variation in a representative autosomal dominant disorder.

Results: We found that in some genes previously reported as important causes of a given cardiomyopathy, rare variation is not clinically informative because there is an unacceptably high likelihood of false-positive interpretation. By contrast, in other genes, we find that diagnostic laboratories may be overly conservative when assessing variant pathogenicity.

Conclusions: We outline improved analytical approaches that evaluate which genes and variant classes are interpretable and propose that these will increase the clinical utility of testing across a range of Mendelian diseases.Genet Med 19 2, 192-203.

Figure 1

Proportion of individuals with rare variants in hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathy (DCM) in combined clinical cohorts (data from Oxford Medical Genetics Laboratory and Laboratory of Molecular Medicine) compared with Exome Aggregation Consortium (ExAC) data (gray columns). Variants counted were single-nucleotide changes or small insertion/deletion variants detected in the coding region ±2 bp, with an ExAC minor allele frequency of <1 × 10⁻⁴ (see Methods). Clinical cohorts: HCM, n = 632 to 6,179 and DCM, n = 121 to 1,315 (see Table 1, Supplementary Table S5a,b online). Information on reported pathogenicity class (red = pathogenic (P), orange = likely pathogenic (LP), yellow = variant of uncertain significance (VUS)) is overlaid. See Supplementary Tables S1, S4a,b online for full details. ^ = genes analyzed in fewer than 200 cases. ExAC: n = mean of total adjusted allele count for rare variant carriers. For HCM genes, n ranged from 47,153 to 60,647; for DCM genes, n was 42,697 to 60,647 (see Supplementary Table S5a,b online). CTF1 and RBM20 were removed from analysis due to poor coverage in ExAC.

Figure 2

Odds ratios (OR) with 95% confidence interval (CI) for each gene tested in the hypertrophic cardiomyopathy (HCM) (n = 632 to 6,179), dilated cardiomyopathy (DCM) (n = 121 to 131), and arrhythmogenic right ventricular cardiomyopathy (ARVC) (n= 93 to 361) clinical cohorts compared with Exome Aggregation Consortium (ExAC) reference samples (n = mean of total adjusted allele count for rare variant carriers. for HCM genes, n = 47,153 to 60,647; for DCM genes, n = 42,697 to 60,647; and for ARVC genes, n = 51,126 to 60,218). See Supplementary Table S5a–c online for data used to generate this plot. Data have been plotted (log₁₀ scale) for all protein-altering variants (black) and truncating variants (blue). For truncating variants, OR with 95% CI have been plotted for genes where a statistically significant difference was observed for this variant type on FET. *Statistically significant Fisher's exact test (FET) (P=0.05 with Bonferroni correction; HCM P ≤ 0.0025; DCM P ≤ 0.001; and ARVC P ≤ 0.006.). ^Genes analyzed in fewer than 200 cases. CTF1 and RBM20 were removed from analysis owing to poor coverage in ExAC.

Figure 3

Variant interpretation. (a) Distribution of rare (Exome Aggregation Consortium (ExAC) minor allele frequency <1 × 10⁻⁴) MYH7 missense variants in hypertrophic cardiomyopathy (HCM) (n = 864) and dilated cardiomyopathy (DCM) (n = 69) clinical cohorts and ExAC (n = 816), with the key myosin protein regions highlighted. Using nonrandom mutation cluster analysis (NMC), (refer to Supplementary Methods online), a significant variant cluster (P < 3 × 10⁻¹⁵, false discovery rate (FDR), q < 5 × 10⁻13) was identified between residues 181 and 937 (involving the motor domain, lever arm, and part of the rod) in HCM cases, and depletion in this region and a significant cluster (P < 3 × 10⁻⁸, FDR q < 4 × 10⁻⁵) was identified between residues 1,271 and 1,903 (in the part of the rod that forms the filament backbone) in ExAC samples. The etiological fraction (EF) for a rare MYH7 missense variant identified in a HCM proband ranges from 0.97 in the HCM cluster to 0.67 in the control cluster. Vertical gray bars depict the positions of variants in cohorts; grayscale shows variant density where variants are coincident. An overview of the genetic landscape of HCM (b) and DCM (c) for truncating (blue) and nontruncating (gray) variants, as well as MYH7 missense variants in the clusters identified in (a) (orange, disease cluster; yellow, ExAC control cluster). The case excess (y-axis) is the frequency of rare variation in disease cohorts over and above the frequency in ExAC and indicates the relative importance of the gene and variant class to the genetic etiology of each cardiomyopathy. The etiological fraction (EF) (x-axis) is an estimate of the proportion of affected carriers where the variant caused the disease; it is a measure of the interpretability of variants of this class (see Supplementary Table S4a,b online for full details). This measure is an average of all variants of a given class; some of which will be pathogenic but others will be benign.

Figure 4

Comparison of the number of variants reported as putatively pathogenic in research studies. Hypertrophic cardiomyopathy (HCM) (a) and dilated cardiomyopathy (DCM) (b) in research studies (using generic analysis criteria such as variant class, missense-effect predictions, and variant population frequency in the Exome Sequencing Project) with those predicted as pathogenic by the excess of variation in cases over Exome Aggregation Consortium (ExAC) in each gene. For the HCM study (a), genes are colored according to the cardiac disease for which they are primarily associated, as defined by Lopes et al. Although there is good concordance between the research findings and the ExAC predictions for established HCM genes, for genes primarily associated with DCM, ARVC, and arrhythmias, the variation in cases is similar to that in ExAC. In the DCM study^,, (b), variation burden in MYBPC3, SCN5A, and MYH6 is similar between the published research cases and ExAC, suggesting that most variants in these genes are unlikely to cause DCM.