genomewide-eQTL test

When we run genome-wide association tests for identifying eQTLs, we can run analyses for multiple genes at the same time. We found it substantially reduced the total computation time due to reducing the overhead of reading genotypes

step 1 needs to be run once for each trait/gene expression and can is shared by cis-eQTL and genomewide-eQTL test
Same as Step 1 - cis-eQTL for each gene
step 1 can be run independently for each gene, so each gene job can use one CPU.

You may SKIP the following step 1 jobs and download the generated step 1 output from here and store tham to ./output

  for i in {1..100}
  do
  echo $i
  step1prefix=./output/nindep_100_ncell_100_lambda_2_tauIntraSample_0.5_gene_${i}
  /bin/time -o ${step1prefix}.runinfo.txt -v Rscript step1_fitNULLGLMM_qtl.R \
      --useSparseGRMtoFitNULL=FALSE  \
      --useGRMtoFitNULL=FALSE \
      --phenoFile=./input/seed_1_100_nindep_100_ncell_100_lambda_2_tauIntraSample_0.5_Poisson.txt \
      --phenoCol=gene_${i}       \
      --covarColList=X1,X2,pf1,pf2    \
      --sampleCovarColList=X1,X2      \
      --sampleIDColinphenoFile=IND_ID \
      --traitType=count \
      --outputPrefix=${step1prefix}   \
      --skipVarianceRatioEstimation=FALSE  \
      --isRemoveZerosinPheno=FALSE \
      --isCovariateOffset=FALSE  \
      --isCovariateTransform=TRUE  \
      --skipModelFitting=FALSE  \
      --tol=0.00001   \
      --plinkFile=./input/n.indep_100_n.cell_1_01.step1       \
      --IsOverwriteVarianceRatioFile=TRUE &> ${step1prefix}.log
  done

Table of contents