Table 1 Summary of existing methods for multi-trait association testing using GWAS summary statistics
Method | Description | Output | Reference (PMID) |
|---|---|---|---|
metaCCA | Perform canonical correlation analysis (CCA) using correlation matrices estimated from summary statistics and reference panel | P-value for association with a set of traits (global association) | 27153689 |
MOSTest | Mahalanobis norm of the vector of z-statistics (\({z}^{T}{\Sigma }^{-1}z\)) with correlation matrix \(\Sigma\) estimated form randomly permuted genotypes | 32665545 | |
JASS | Omnibus test statistic based on \({z}^{T}{\Sigma }^{-1}z\) and sumZ statistic \(\frac{{\left({w}^{T}Z\right)}^{2}}{{w}^{T}\Sigma w}\) | 32002517 | |
MultiPhen | Regression with the genotype as dependent variable and phenotype for multiple traits as independent variable | 22567092 | |
metaMANOVA | Test association using multivariate analysis of variance statistic, highly similar to MOSTest and JASS Omnibus test. Correlation matrix \(\Sigma\) is estimated using SNPs with no association with the traits. | 29226385 | |
metaUSAT | Optimal combination of metaMANOVA and sum of squared score (SSU) statistics | 29226385 | |
HIPO | Search for the linear combination of multi-trait summary statistics that maximizes average non-centrality parameter across SNPs | 30289880 | |
MTAG | Use multi-trait summary statistics to obtain single-trait effect size estimates by incorporating a prior distribution on the effect size | Estimate of individual-trait GWAS effect size and associated test statistic | 29292387 |
ASSET/fastASSET | Search for optimal subset of traits that maximizes meta-analysis z-statistic | P-value for association with a set of traits (global association) and a subset of selected traits | 22560090 |
