Robust-joint-interaction by epstein-software

Robust Joint Tests of SNP and SNP-Environment Interaction

Introduction

In Almli et al. (2014) we created a robust test of SNP and SNP-environment interaction for complex traits that used Huber-White estimates of variance to eliminate bias arising from heteroscedasticity or other model misspecification. We implemented this test in R script called with PLINK.

We illustrate how to use the R function below.

The R base system

R is a widely-used, free and open source software environment for statistical computing and graphics. The most recent version of R can be downloaded from the Comprehensive R Archive Network (CRAN) CRAN provides precompiled binary versions of R for Windows, MacOS, and select Linux distributions that are likely sufficient for many users' needs. Users can also install R from source code; however this may require a significant amount of effort. For specific details on how to compile, install, and manage R and R-packages, refer to the manual R Installation and Administration.

R packages required for analysis

Our robust test of SNP and SNP-environment interaction for complex traits requires the installation of two R libraries:

Rserve, required for PLINK to run the R plugin function and
geepack, required to obtain the robust variance estimators for the joint interaction test.

The easiest method to install these packages is to open a R session and enter the command:

install.packages("geepack")

Choose a mirror and the package should automatically download.

A similar command,

install.packages("Rserve")

should install the Rserve package.

One can also install R packages from the command line see more details here.

Please note that PLINK only provides support for R functions in their Linux-based and Mac OS distributions.

Starting Rserve

There are a few ways to start Rserve in daemon mode. I. One method uses the command line; in particular: R CMD Rserve

Figure 1. Terminal output after starting Rserve from command line.

II. Alternatively, one can start Rserve by saving the following two lines into a script file called run-Rserve.R:

library(Rserve)<br>
Rserve(args="--no-save")

Then the script can be executed through R from the command line with command R --vanilla < run-Rserve.R.

As well, these two lines of R code could also be executed within an R console or an R environment such as RStudio.

Using PLINK to perform robust joint test

Once Rserve is installed and running locally, we can use PLINK to invoke the R function to perform robust joint testing of SNP and SNP-environment interaction. We provide sample PLINK files (PED/MAP/COV) consisting of 500 individuals genotyped for 10 SNPs and assayed for three environmental predictors. For each of the 10 SNPs, we wish to perform a joint test and SNP and SNP-environment interaction using the first environmental predictor listed in the COV file (the second and third environmental predictors in the COV file are treated only as covariates). To perform these analyses on our test dataset, we invoke the following PLINK command:

plink --file test-int-plink --covar test-int-plink.cov --R robust-joint-int-plugin.R

The command will generate a file 'plink.auto.R'. Each row of the file corresponds to results from a tested SNP and has the following fields (in order):

Chromosome position
SNP id
Physical Position (base-pair)
Minor allele
P-value of model-based test of SNP and SNP-environment interaction
P-value of robust test of SNP and SNP-environment interaction

Using the provided sample data, the contents of the this file is shown here:

   1    SNP1     101000    2 0.864095   0.839595    
   1    SNP2     102000    2 0.776515   0.836906    
   1    SNP3     103000    2 0.386562   0.376118    
   1    SNP4     104000    2 0.532115   0.642094    
   1    SNP5     105000    2 3.47487e-09    5.04597e-06 
   1    SNP6     106000    2 0.543536   0.563148    
   1    SNP7     107000    2 0.451066   0.397487    
   1    SNP8     108000    2 0.758953   0.75529 
   1    SNP9     109000    2 0.94371    0.947891    
   1   SNP10     110000    2 0.972809   0.972334

The R code can easily be modified to consider other joint interaction tests of interest. Also, bed, bim, and fam files can also be used.

Important: if one wishes to adjust the robust joint test for gender, one must include a numeric gender variable in the PLINK COV file. Use of the --sex flag in PLINK will not adjust the analysis specified in the R plugin function for gender.

More details on writing, calling, and debugging R functions in PLINK can be found in the PLINK documentation for R plugin functions.

Questions and Technical Support

For questions or concerns with the R function, please contact Richard Duncan and Michael Epstein

We appreciate any feedback you have with our site and instructions.

Robust-joint-interaction