raydebashree / usat Goto Github PK

USAT (Unified Score-based Association Test) uses a data-adaptive weighted score-based test statistic for testing association of multiple continuous phenotypes with a single genetic marker. The R function usat implements this association test. For details of this statistical method, please refer/cite:

Ray, D., Pankow, J.S., Basu, S. "USAT: A Unified Score-based Association Test for Multiple Phenotype-Genotype Analysis". Genetic Epidemiology, 40(1):20-34, 2016. PMID: 26638693

Key Words: GWAS; MANOVA; Multiple phenotypes; Multivariate analysis; Pleiotropy; Score test

R (>= 3.0.1), CompQuadForm, minqa, survey

require(devtools)
source_url("https://github.com/RayDebashree/USAT/blob/master/usat_v1.21.R?raw=TRUE")

It is recommended to download/copy the stand-alone R program in this repository, save it in your local directory of choice and source() it from your local directory. When a new version of the software is available, older versions may be removed from this repository, and the above devtools::source_url() technique may not work.

Version 1.21 - December 12, 2016

An updated version of the software with more user controls.

Version 1.1 - April 07, 2016

First public release of the software.

usat(Y, X, COV=NULL, na.check=TRUE, na.check.msg=TRUE, manova.out=FALSE, AbsTol=.Machine$double.epsˆ0.8)

source("usat_v1.21.R")

# simulate 2 phenotypes on 1000 individuals
library(MASS) # needed for multivariate normal simulation
Y<-mvrnorm(n=1000, mu=c(0,0), Sigma=matrix(c(1,0.2,0.2,1),2,2))

# simulate a single marker for 1000 individuals
X<-matrix(rbinom(n=1000, size=2, prob=0.2), ncol=1) # additive model

# apply USAT to test association
u.out<-usat(Y=Y, X=X, COV=NULL, na.check=FALSE)

# USAT test statistic and p-value
t<-u.out$T.usat
p<-u.out$p.usat

1. The method USAT and its software is designed for multiple continuous phenotypes from a random sample. If the ascertainment of individuals in the sample is non-random (e.g., in case-control retrospective study design), it is advisable to account for the sampling scheme (e.g., adjusting the sampling variable as a covariate) when using USAT. One may also use methods and tools designed specifically for the analysis of secondary phenotypes. We proposed one such method (POM-PS) and its software is publicly available.

   • Caution: USAT is a single-variant association test; so not expected to work well for rare variants (i.e., genetic variants with very low allele-frequencies).

2. Although USAT software can adjust for covariates, it is advisable to perform covariate adjustments and necessary transformations (e.g., inverse-normal transformation) on the traits and then apply USAT on the residuals. This will speed up total computation time.

3. The method USAT and its software is designed for unrelated individuals. If you have two cohorts with overlapping samples and you want to analyse the combined sample, it is desirable to exclude the overlapping individuals, and any related individuals.

4. If you receive an error like the integral is probably divergent, try reducing the absolute tolerance parameter AbsTol.

5. If you wish to test genetic association of multiple traits (categorical and/or continuous) and you have access to summary statistics only, please use the new software metaUSAT that is publicly available now!

6. If removal of related individuals is a concern (e.g., sample size gets greatly reduced) or if there is cryptic relatedness that needs to be accounted for, it is advisable to first obtain the GWAS summary statistics for each trait by analyzing each trait separately accounting for the relatedness among the individuals (e.g., using EPACTS to analyze each trait). Then apply metaUSAT on the single trait GWAS summary statistics.

7. USAT, like most other cross-phenotype association methods, assumes multivariate normality of the underlying traits. While transformations like rank-based inverse-normalization help induce univariate normality of each trait, it does not guarantee multivariate normality of the joint distribution of the traits. More on this coming up soon!