HCLC-FC

HCLC-FC performs multivariate association testing for phenome-wide association studies by clustering phenotypes, combining cluster-level test statistics, and controlling the false discovery rate to detect genetic associations across phenotype categories.

Key Features:

• Hierarchical Clustering: Bottom-up hierarchical clustering partitions large numbers of phenotypes into disjoint clusters within each phenotypic category based on genetic architecture across phenotypes.
• Clustering Linear Combination: A clustering linear combination method integrates test statistics from phenotypic clusters to generate p-values reflecting associations within each phenotypic category.
• False Discovery Rate Control: A novel false discovery rate (FDR) control approach is applied to maintain rigorous control of false positives across multivariate tests.
• Simulation Validation: Extensive simulation studies demonstrate control of FDR at nominal levels and improved power relative to existing methods.
• Empirical SNP Discovery: Application to over 300,000 UK Biobank samples identified 1,244 significant single nucleotide polymorphisms (SNPs) reported in the GWAS catalog.

Scientific Applications:

• EMR-based PheWAS: Analysis of electronic medical record (EMR) phenotypes to detect genotype-phenotype associations across many correlated traits.
• Large-cohort genetic association studies: Multivariate testing in large datasets such as the UK Biobank to improve detection of genetic signals across phenotype categories.
• SNP discovery and cataloging: Identification and reporting of significant SNPs for inclusion in resources like the GWAS catalog.

Methodology:

Bottom-up hierarchical clustering to partition phenotypes; clustering linear combination to integrate cluster-level test statistics and produce p-values per phenotypic category; a novel FDR control procedure; extensive simulation studies confirming FDR control and power; applied to over 300,000 UK Biobank samples yielding 1,244 significant SNPs reported in the GWAS catalog.