Background: While next-generation sequencing has enabled us to rapidly identify sequence variants, clinical application is limited by our ability to determine which rare variants impact disease risk. Aim: Developing computational methods to identify clinically important variants Methods and Results: (1) I built a disease-specific variant classifier for inherited cardiac conditions (ICCs), which outperforms genome-wide tools in a wide range of benchmarking. It discriminates pathogenic variants from benign variants with global accuracy improved by 4-24% over existing tools. Variants classified with >90% confidence are significantly associated with both disease status and clinical outcomes. (2) To better interpret missense variants, I examined evolutionarily equivalent residues across protein domain families, to identify positions intolerant of variations. Homologous residue constraint is a strong predictor of variant pathogenicity. It can identify a subset of de novo missense variants with comparable impact on developmental disorders as protein-truncating variants. Independent from existing approaches, it can also improve the prioritisation of disease-relevant gene for both developmental disorders and inherited hypertrophic cardiomyopathy. (3) TTN-truncating variants are known to cause dilated cardiomyopathy, but the effect of missense variants is poorly understood. Using the approach in (2), I studied the role of TTN missense variants on DCM. Our prioritised residues are enriched with known pathogenic variants, including the two known to cause DCM and others involved in skeletal myopathies. I also found a significant association between constrained variants of TTN I-set domains and DCM in a case-control burden test of Caucasian samples (OR=3.2, 95%CI=1.3-9.4). Within subsets of DCM, the association is replicated in alcoholic cardiomyopathy. (4) Finally, I also developed a tool to annotate 5'UTR variants creating or disrupting upstream open reading frames (uORF). Its utility is demonstrated to detect high-impact uORF-disturbing variants from ClinVar, gnomAD and Genomics England. Conclusion: These studies established broadly applicable methods and improved understanding of ICCs.