How Functional Genomics Can Keep Pace With VUS Identification

Figures

• 

  Figure 1. Cardiac disease proteins highlighted in this review. (A) Arrhythmias can result from structurally abnormal hearts (e.g., DCM) or normal hearts with electrical abnormalities (e.g., LQTS). (B) Simplified cartoon highlighting each protein and their primary functions. (C) Ventricular action potential showing the major ionic currents. (D) Graph illustrating the range in size for each protein discussed on the x-axis from small calmodulin to the the largest human protein titin comprised of hundreds of small soluble domains (eg., IgD). The y-axis shows the relative pleiotropy for each gene (i.e., relative number of diseases associated with each). For example, CaM has only been linked to LQTS while lamin underlies over a dozen diseases. Nav1.5 on the other hand is a good example of pleiotropy among the ion channels reviewed by Cerrone et al. (3) with lamin being the most pleiotropic. DCM, dilated cardiomyopathy; SR, sarcoplasmic reticulum; RYR2, ryanodine receptor 2; CaM, calmodulin; CCD, coiled-coil domain; CD, converter domain; IgD, immunoglobulin-like domain; NTD, FN-3, fibronectin type III-like; N-terminal domain; TMD, transmembrane domain; PASD, PerArntSim domain; CNBD, cyclic…

• 

  Figure 2. Genes associated with inherited cardiac arrhythmias. (A) The total number of variants for each gene in ClinVar and their classifications [conflicting interpretation (CI), likely benign (LB), benign (B), variant of uncertain significance (VUS), likely pathogenic (LP) and pathogenic (P)] as of June 2022. Inset shows relative percentages for all genes combined. Disease associations are not comprehensive (e.g., LMNA is linked to many non-cardiac diseases). Genes reviewed here shaded gray. (B) Graph illustrating the range of effects variants can have toward pathogenicity. Figure adapted from Ingles et al. (8). Disease abbreviations: CPVT, catecholaminergic polymorphic ventricular tachycardia; HCM, hypertrophic cardiomyopathy; RCM, restrictive cardiomyopathy; ARVC, arrhythmogenic ventricular cardiomyopathy; BrS, Brugada syndrome; LQTS, long QT syndrome; DCM, dilated cardiomyopathy; CCD, cardiac conduction disease; SIDS, sudden infant death syndrome; SQTS, short QT syndrome; TS, Timothy syndrome; JLNS, Jeverell and Lange-Nielsen syndrome; ATS, Andersen-Tawil syndrome; LVNC, left ventricular non-compaction cardiomyopathy.

• 

  Figure 3. AMCG Guidelines for variant interpretation. A brief overview of the eight lines of supporting evidence for evaluating variants of uncertain significance. Numerous benign and pathogenic supporting computational and functional levels are shown for each approach with functional evidence filled blue. The lines of evidence described by Richards et al. (27) in the guidelines are as follows: BS3/PS3, not damaging/damaging functional studies; BP1, missense variant in a gene with primarily truncating variants; BP3, indels in repeat region without known function; BP4/PP3, multiple lines of computational evidence showing no damaging effects/damaging affects; BP7, non-splicing, silent variant; PP2, variant in gene where pathogenic variants are common and benign variants are few; PM1, variant hotspot without benign variants; PM4, protein length changing variant; PM5, variant at residue with other pathogenic variants; PS1, amino acid change same as other pathogenic variant(s); PVS1, null variant in a pathogenic gene with known LOF.

• 

  Figure 4. Functional and computational methods of inherited cardiac arrhythmias. (A) Animal and iPS-CM functional models will remain important but are low throughput. (B) Overexpression systems are higher throughput especially when automated with functional results for ~100 variants reported in less than two months' time by Ng et al. (38). (C) Multiple assays of variant effects (MAVEs), while largely proxy assays, are magnitudes higher throughput allowing for comprehensive variant analyses. (D) Tools to predict a VUS can be simple and rule-based relying on sequence and/or structural data or, more commonly, machine learning methods are employed. FACS, fluorescence-activated cell sorting; HGMD, Human Gene Mutation Database; UniProt, Universal Protein Resource; genomAD, Genome Aggregation Database; REVEL, Rare Exome Variant Ensemble Learner; dbNSFP, Database for Functional Predictions of non-synonymous SNPs.