PanGEMMA: Genome-wide Efficient Mixed Model Association

This repository is used to rewrite and modernize the original GEMMA tool. The idea it to upgrade the software, but keeping it going using ideas from Hanson and Sussman's book on Software Design for Flexibility: How to Avoid Programming Yourself into a Corner. It is work in progress (WIP). For more information see PanGEMMA design

GEMMA is the original software toolkit for fast application of linear mixed models (LMMs) and related models to genome-wide association studies (GWAS) and other large-scale data sets. You can find the original code on github.

Check out RELEASE-NOTES.md to see what's new in each release.

Key features
Installation
Precompiled binaries
Run GEMMA
Debugging and optimization
Help
Citing GEMMA
License
Optimizing performance
Building from source
Input data formats
Reporting a GEMMA bug or issue
Check list:
Code of conduct
Credits

Key features

Fast assocation tests implemented using the univariate linear mixed model (LMM). In GWAS, this can correct for population structure and sample non-exchangeability. It also provides estimates of the proportion of variance in phenotypes explained by available genotypes (PVE), often called "chip heritability" or "SNP heritability".
Fast association tests for multiple phenotypes implemented using a multivariate linear mixed model (mvLMM). In GWAS, this can correct for population structure and sample (non)exchangeability - jointly in multiple complex phenotypes.
Bayesian sparse linear mixed model (BSLMM) for estimating PVE, phenotype prediction, and multi-marker modeling in GWAS.
Estimation of variance components ("chip/SNP heritability") partitioned by different SNP functional categories from raw (individual-level) data or summary data. For raw data, HE regression or the REML AI algorithm can be used to estimate variance components when individual-level data are available. For summary data, GEMMA uses the MQS algorithm to estimate variance components.

Installation

WIP

Precompiled binaries

WIP

Run GEMMA

GEMMA is run from the command line. To run gemma

gemma -h

a typical example would be

# compute Kinship matrix
gemma -g ./example/mouse_hs1940.geno.txt.gz -p ./example/mouse_hs1940.pheno.txt \
    -gk -o mouse_hs1940
# run univariate LMM
gemma -g ./example/mouse_hs1940.geno.txt.gz \
    -p ./example/mouse_hs1940.pheno.txt -n 1 -a ./example/mouse_hs1940.anno.txt \
    -k ./output/mouse_hs1940.cXX.txt -lmm -o mouse_hs1940_CD8_lmm

Above example files are in the git repo and can be downloaded from github.

Debugging and optimization

GEMMA has a wide range of debugging options which can be viewed with

 DEBUG OPTIONS

 -check                   enable checks (slower)
 -no-fpe-check            disable hardware floating point checking
 -strict                  strict mode will stop when there is a problem
 -silence                 silent terminal display
 -debug                   debug output
 -debug-data              debug data output
 -debug-dump              -debug-data, but store the data to files (grep write() calls for messages/names)
 -nind       [num]        read up to num individuals
 -issue      [num]        enable tests relevant to issue tracker
 -legacy                  run gemma in legacy mode

typically when running gemma you should use -debug which includes relevant checks. When compiled for debugging the debug version of GEMMA gives more information.

For performance you may want to use the -no-check option. Also check the build optimization notes in INSTALL.md.

Help

Citing PanGEMMA

PanGEMMA is not published yet.

But if you use GEMMA for published work, please cite our paper:

Xiang Zhou and Matthew Stephens (2012). Genome-wide efficient mixed-model analysis for association studies. Nature Genetics 44, 821–824.

If you use the multivariate linear mixed model (mvLMM) in your research, please cite:

Xiang Zhou and Matthew Stephens (2014). Efficient multivariate linear mixed model algorithms for genome-wide association studies. Nature Methods 11, 407–409.

If you use the Bayesian sparse linear mixed model (BSLMM), please cite:

Xiang Zhou, Peter Carbonetto and Matthew Stephens (2013). Polygenic modeling with bayesian sparse linear mixed models. PLoS Genetics 9, e1003264.

And if you use of the variance component estimation using summary statistics, please cite:

Xiang Zhou (2016). A unified framework for variance component estimation with summary statistics in genome-wide association studies. Annals of Applied Statistics, in press.

License

The PanGEMMA and GEMMA source code repository is free software: you can redistribute it under the terms of the GNU General Public License. All the files in this project are part of GEMMA. This project is distributed in the hope that it will be useful, but without any warranty; without even the implied warranty of merchantability or fitness for a particular purpose. See file LICENSE for the full text of the license.

Both the source code for the gzstream zlib wrapper and shUnit2 unit testing framework included in GEMMA are distributed under the GNU Lesser General Public License, either version 2.1 of the License, or (at your option) any later revision.

The source code for the included Catch unit testing framework is distributed under the Boost Software Licence version 1.

The GEMMA software was developed by:

Xiang Zhou
Dept. of Biostatistics
University of Michigan

with contributions from Peter Carbonetto, Tim Flutre, Matthew Stephens, and others.