MetaSKAT

MetaSKAT performs meta-analysis of gene- or region-based multimarker rare variant association tests in sequencing association studies by aggregating study-level score statistics and between-variant covariance (linkage disequilibrium, LD) summaries to increase power and accommodate heterogeneity across studies.

Key Features:

• Statistical framework: A general framework for meta-analysis of gene- or region-based multimarker rare variant tests, including burden tests and variance component tests.
• Score-statistics based aggregation: Uses study-specific score statistics that require fitting only a null model per study, avoiding unstable regression coefficient estimates for rare variants.
• Between-variant covariance (LD) summaries: Incorporates covariance-type statistics (linkage disequilibrium) for each gene or region using study-specific summary statistics.
• Heterogeneity handling: Accommodates varying levels of genetic effect heterogeneity across studies, including analyses across multiple ancestry groups.
• Comparable power to pooled analyses: Demonstrates power equivalent to joint analyses that pool individual-level genotype data without requiring individual-level pooling.
• Extensive validation: Performance evaluated through extensive simulations under varying levels of heterogeneity across studies.

Scientific Applications:

• Sequencing association study meta-analysis: Meta-analysis of rare variant associations using gene- or region-based multimarker tests across multiple studies.
• Multi-ancestry meta-analyses: Analyses that combine studies from diverse ancestry groups while accounting for heterogeneity in genetic effects.
• Trait association studies: Application to uncover rare variant contributions to traits such as blood lipid levels in multicohort sequencing studies.

Methodology:

For each study fit a null model and compute study-specific score statistics and between-variant covariance (LD) matrices per gene/region, then aggregate the score statistics across studies within a general framework for burden and variance component tests; performance assessed by simulations under varying heterogeneity.