metaCCA

metaCCA performs multivariate association analysis of GWAS summary statistics to detect joint genotype–phenotype correlations when individual-level genotype–phenotype data are unavailable.

Key Features:

• Multivariate representation: Enables simultaneous analysis of multiple phenotypes and genotypes to detect associations not apparent in univariate tests.
• Extension of canonical correlation analysis (CCA): Adapts traditional CCA to operate on GWAS summary statistics in the absence of individual-level records.
• Covariance shrinkage algorithm: Applies covariance shrinkage to stabilize covariance estimates in high-dimensional GWAS data.
• Summary-statistics-based analysis: Operates on aggregated GWAS summary statistics rather than requiring individual-level genotype or phenotype data.

Scientific Applications:

• Multivariate meta-analysis: Applied to multivariate meta-analysis of genetic studies to combine information across studies.
• NMR metabolomics in Finnish cohorts: Demonstrated on two Finnish studies involving nuclear magnetic resonance metabolomics.
• Comparison with pooled individual-level analyses: Produced results in excellent agreement with pooled individual-level analyses using standard univariate output from SNPTEST.
• High-throughput phenotyping: Used to uncover multivariate signals in high-throughput phenotyping technologies.
• Lipid gene discovery: Detected strong multivariate signals in lipid genes relevant to metabolic trait genetics.

Methodology:

Extends canonical correlation analysis to work on GWAS summary statistics, integrates summary statistics from one or multiple studies, and applies a covariance shrinkage algorithm to stabilize covariance estimation.