Genard

From MezeyLabWiki

Jump to: navigation, search

Genard corrects for genetic confounding due to population structure and kinship in genome-wide association studies. Using a data-adaptive low rank linear mixed model, genard learns the dimensionality of the correction from the data. The software is compatible with plink files and can analyze 650,000 markers for ~2000 individuals in ~15 minutes or ~6000 individuals in 2 hours. The paper describing genard is currently in review.

Download genard [1]

Description

Genard fits a low rank linear mixed model and uses a subset of principal components of the genetic relationship matrix to account for population structure and kinship. Genard uses a metric of model complexity termed "effective degrees of freedom" to determine the rank of the correction and evaluates fit of the model to the data based on AIC, BIC, Generalized Cross-Validation (GCV) or the log-likelihood.

Running genard

Generate a genetic relationship matrix from a set of markers using EMMAX [2], GEMMA [3], FaST-LMM [4], or other software.

See options for genard by entering command.

genard

This prints:

#----------------------------------------------------------#
|       genard        |      v1.0      |    May 17 2013    |
|----------------------------------------------------------|
|         (C) 2013 Mezey Lab @ Cornell University          |
|----------------------------------------------------------|
|           http://mezeylab.cb.bscb.cornell.edu/         |
#----------------------------------------------------------#

Parameters: 
genard 

--method must be specified

Commandline parameters
--method         [SMA | LMMRS]
--criterion      [AIC | BIC | GCV | logLik]
--order      	 [STD | COR | CORLAMBDA | DF]
--maxRank        [LMM Rank Search for rank 0..maxRank]
--tped           [genotype data in plink TPED format]
--tfam           [phenotype (and sex) data in plink TFAM format]
--sex            [if --tfam is used, this includes sex as a covariate]
--grm            [genetic relationship matrix (GRM)]
--saveEigenDecomp[save eigen decomposition of GRM]
--eigenDecomp    [read eigen decomposition of GRM]
--covariates     [file storing matrix with each column being a covariate]
--regression     [specify regression model as either LINEAR or LOGISTIC]
                 (default = LINEAR)
--name           [name to be appended to results files]
--nthreads       [number of threads used to run in parallel]
                 (default = machine default)

Genard can perform two types of analysis:

  • Single marker analysis (SMA) is a fixed effects regression model with covariates determined using the --covariates flag. This method is proivded for performing standard analysis for either linear or logistic models.
 genard --method SMA --tped test.tped --tfam test.tfam
  • Linear mixed model with rank search (LMMRS) fits a low rank linear mixed model with a genetic relationship matrix where the rank is learned from the data. Relevant flags are --criterion, --order, --grm and --maxRank.

Run an analysis where principal components are ordered by eigen-value (i.e. LAMBDA), the rank is determined by the best BIC, and the maximum rank considered is 400:

genard --method LMMRS --tped test.tped --tfam test.tfam --grm K.txt --order LAMBDA --criterion BIC --maxRank 400 --name try1

Other combinations for --order and --criterion can be used based on the options shown above.

Personal tools