meta_tissue: Ultra-fast meta-tissue algorithm

In brandonjew/mcLMM: Ultra-fast association testing across multiple contexts with linear mixed models

Description

Provides identical results as meta-tissue in linear time with respect to the number of samples rather than cubic. Slight differences may be due to different optimization algorithms of the same likelihood function.

Usage

meta_tissue(
  expr,
  geno,
  covs = NULL,
  heuristic = FALSE,
  newRE = TRUE,
  force.iter = FALSE,
  verbose = TRUE
)

Arguments

expr	Matrix with individuals as rows and tissues as columns. Missing gene expression must be NA
geno	Vector of genotypes for each individual
covs	Matrix with individuals as rows and covariates as columns.
heuristic	Boolean. Uses heuristic for scaling standard errors. increases sensitivity at the cost of higher FPR.
newRE	Boolean. Use new random effects model to perform meta analyses as discussed in Sul et al.
force.iter	Boolean. If TRUE, force iterative method even when there is no missing data. This is included for testing purposes only. The optimal non-iterative method for no missing data is exact and way faster.
verbose	Boolean. Output logging info.

Value

List of estimated coefficients beta, coefficient correlation corr, and sigma_g.

brandonjew/mcLMM documentation built on Dec. 19, 2021, 11:42 a.m.