Abstract Body (Do not enter title and authors here): Background: Identifying novel etiologies for congenital heart disease (CHD) remains challenging due to substantial genetic heterogeneity, variable expressivity, and incomplete penetrance. Current standard guidelines for variant assessment (e.g., ACMG/AMP criteria) were originally developed for highly penetrant monogenic disorders with Mendelian inheritance patterns, limiting their effectiveness for conditions like CHD that frequently exhibit complex, oligogenic, or multifactorial inheritance. This study aimed to develop a variant prioritization and gene identification strategy specifically tailored to address the genetic intricacies of CHD.

Methods: We analyzed genomic and phenotypic data from 337 CHD probands from three independent cohorts (DDD, PCGC, Nationwide Children’s Hospital), each with confirmed CHD diagnoses and exome/genome sequencing. We trained ChiVaLRi, a supervised machine-learning classifier, to prioritize causal variants using variant pathogenicity, phenotype concordance, and embryonic heart RNA-seq expression data. ChiVaLRi also incorporates an RNA-seq-based module that predicts novel CHD gene involvement through gene co-expression with known cardiac genes. Classifier performance was assessed with rank-based metrics and precision-recall curves, benchmarked against leading phenotype-aware gene-prioritization tools developed for Mendelian disorders: Exomiser, LIRICAL, AI-MAVERICK, and MARRVEL.

Results: ChiVaLRi ranked the causal variant among the top five candidates in nearly 90% of CHD cases, significantly outperforming existing tools (p < 0.001). ChiVaLRi also highlighted plausible candidate variants in previously unsolved cases and accurately identified recently validated CHD genes. The highest-scoring novel gene candidates predicted by ChiVaLRi were significantly enriched in cardiac developmental pathways (p < 0.01).

Conclusions: ChiVaLRi provides a robust, cardiac-specific variant prioritization strategy that accelerates genetic diagnosis, improves clinical efficiency, and supports personalized care for CHD patients. Additionally, its gene-prediction module facilitates the discovery of novel CHD-associated genes, advancing insights into cardiac developmental biology. ChiVaLRi is freely available at https://github.com/nch-cloud/ods-chivalri.