CardioClassifier: disease- and gene-specific computational decision support for clinical genome interpretation

Abstract

Purpose: Internationally adopted variant interpretation guidelines from the American College of Medical Genetics and Genomics (ACMG) are generic and require disease-specific refinement. Here we developed CardioClassifier ( http://www.cardioclassifier.org ), a semiautomated decision-support tool for inherited cardiac conditions (ICCs).

Methods: CardioClassifier integrates data retrieved from multiple sources with user-input case-specific information, through an interactive interface, to support variant interpretation. Combining disease- and gene-specific knowledge with variant observations in large cohorts of cases and controls, we refined 14 computational ACMG criteria and created three ICC-specific rules.

Results: We benchmarked CardioClassifier on 57 expertly curated variants and show full retrieval of all computational data, concordantly activating 87.3% of rules. A generic annotation tool identified fewer than half as many clinically actionable variants (64/219 vs. 156/219, Fisher's P = 1.1 × 10^-18), with important false positives, illustrating the critical importance of disease and gene-specific annotations. CardioClassifier identified putatively disease-causing variants in 33.7% of 327 cardiomyopathy cases, comparable with leading ICC laboratories. Through addition of manually curated data, variants found in over 40% of cardiomyopathy cases are fully annotated, without requiring additional user-input data.

Conclusion: CardioClassifier is an ICC-specific decision-support tool that integrates expertly curated computational annotations with case-specific data to generate fast, reproducible, and interactive variant pathogenicity reports, according to best practice guidelines.

Keywords: bioinformatics; clinical genomics; inherited cardiac conditions; next-generation sequencing; variant interpretation.

Figure 1. Example variant report output by CardioClassifier.

A grid is output for each individual variant. Rules highlighted in colour are activated for the variant and rules in grey on a white background are assessed but not activated. A user can click on a rule to manually add or remove a piece of evidence. All evidence used to assess the variant is displayed under the grid along with links out to external resources. An overall classification for the variant using the ACMG/AMP logic is displayed in the top left corner. *EF - etiological fraction; the prior probability that a variant, identified in a case, is Pathogenic.

Figure 2. Examples of disease-specific optimisation of ACMG/AMP rules.

(a) Missense variants within a sub-portion of MYH7, when identified in a HCM patient, have a 97% prior probability of being Pathogenic (etiological fraction; EF=0.97). We activate PM1 for missense variants in this region. Here we use MYH7:c.2221G>T as an example (labelled with a black bar). (b) Truncating variants in TTN are only known to cause DCM when found in exons constitutively expressed in the heart (proportion spliced in (PSI) > 0.9). We activate PVS1_strong for these variants. Here we use TTN:c.86641delC as an example (labelled with a black bar). (c) Variants that have been identified as Pathogenic in paralogous genes may identify residues that are intolerant to variation. We have created two modified rules, PS1_moderate and PM5_supporting to incorporate this evidence. Here we use KCNQ1:p.T311I as an example. KCNQ2:p.T276I is associated with Ohtahara syndrome. We activate PS1_moderate for KCNQ1:p.T311I which is the equivalent missense change (i.e. same reference and alternate amino acids) in a different member of the same protein family.

Figure 3. Validation of CardioClassifier

(a) Comparing CardioClassifier to a set of 57 MYH7 expert panel curated variants. Rules were split into those that can be computationally annotated and those that are 'case-level' and require manual input. CardioClassiifer was run using an 'All Cardiomyopathy' test to reflect the spectrum of phenotypes caused by variants in MYH7. *Of the computational rules, 3 were removed from the comparison as they represent draft modifications to the ACMG framework by the ClinGen Cardiovascular domain working group that were not published at the time of this work, and not yet implemented in CardioClassifier. Specifically, truncating variants in MYH7 activate a new rule PVS1_moderate. Additionally, for variants classified as Benign by frequency alone (BA1) CardioClassifier does not assess any further rules, leading us to remove an additional data point from the comparison as we would not expect it to be retrieved. (b) Counts of individual rules activated by CardioClassifier and InterVar for 219 variants identified as Pathogenic or Likely Pathogenic in ClinVar. Only pathogenic evidence rules and rules activated by one of the tools at least once are shown.