MCA

MCA compares multilocus SNP-set methods using summary statistics from genome-wide association studies (GWAS) to evaluate type I error control, statistical power across genetic architectures (including polygenic and sparse), and scalability for multilocus association analysis of common and rare variants.

Key Features:

• Comprehensive Comparison: Empirically compares 22 commonly-used summary-statistics-based SNP-set methods for assessment of type I error rates and statistical power.
• Methodological Diversity: Implements burden tests, score-based variance component tests such as the sequence kernel association test, and linkage-disequilibrium-free P value combination methods including the harmonic mean P value method and the aggregated Cauchy association test.
• Performance Evaluation: Evaluates power under diverse simulation scenarios and reports that burden tests are often underpowered while score-based variance component tests show substantial power in polygenic architectures for both common and rare variants.
• Sparse Genetic Architecture Analysis: Identifies two linkage-disequilibrium-free P value combination methods as superior in simulations and real-data applications, including expression quantitative trait loci (eQTL)-weighted integrative analysis.
• Scalability and Computational Efficiency: Assesses computational efficiency and scalability for biobank-scale datasets and reports variable runtimes across included methods.

Scientific Applications:

• Post-GWAS Analysis: Guides selection among multilocus association methods for downstream analysis of GWAS summary statistics.
• Complex Trait Aggregation: Aggregates SNP data within genes to complement single-marker analyses and provide biologically meaningful insight into complex traits.
• Variant Association Studies: Applies to both common and rare variant association studies to identify genetic associations with phenotypic traits.
• eQTL-weighted Integrative Analysis: Supports integrative analyses that weight SNP-set tests by expression quantitative trait loci data to improve detection in sparse architectures.

Methodology:

Implements and empirically compares 22 summary-statistics-based SNP-set methods (including burden tests, score-based variance component tests such as the sequence kernel association test, and linkage-disequilibrium-free P value combination methods like the harmonic mean P value and aggregated Cauchy association test) and evaluates type I error control, statistical power via simulation and real-data scenarios (including eQTL-weighted analysis), and computational efficiency for large-scale datasets.