ClusterBurden

ClusterBurden integrates gene burden and amino acid residue position to improve prediction of gene- and missense rare-variant pathogenicity in case-control studies of protein-coding regions such as Mendelian diseases.

Key Features:

• Joint Analysis Framework: integrates gene burden with amino acid residue position to jointly evaluate gene- and missense-variant pathogenicity.
• BIN-test for Clustering Detection: implements the BIN-test to detect missense variant clustering within protein regions, with simulations showing higher power than Anderson-Darling and Kolmogorov-Smirnov tests.
• Generalized Additive Modeling (GAM): uses GAMs to identify mutational hotspots by integrating burden and clustering information and to refine regional burden maps.
• Hotspot Models: provides hotspot and hotspot+ models that leverage burden, clustering information, and in silico predictors to assess mutational hotspots.
• Performance and Efficiency: demonstrated increased computational efficiency and statistical power in analyses of hypertrophic cardiomyopathy cohorts, in some cases equivalent to increasing sample size by 50% for genes with strong clustering signals.
• Integration with ACMG Criteria: hotspot+ models integrate ACMG criteria PM1 and PP3 to provide evidence for pathogenicity and support reclassification of variants of uncertain significance.

Scientific Applications:

• Mendelian disease gene discovery: increases power to detect disease-associated genes in case-control studies of rare missense variants that cluster within protein regions.
• Variant interpretation and reclassification: refines pathogenicity assessments by combining positional, burden, and in silico evidence and supports application of ACMG PM1 and PP3.
• Hypertrophic cardiomyopathy analyses: applied to HCM cohorts to identify gene-specific clustering and mutational hotspots.

Methodology:

Performs a joint analysis of gene burden and amino acid residue position; applies the BIN-test to detect missense clustering and compares its power to Anderson-Darling and Kolmogorov-Smirnov in simulations; fits generalized additive models to integrate burden and clustering and to define hotspot and hotspot+ models that incorporate in silico predictors and integrate ACMG PM1 and PP3.