Post date: Aug 27, 2014 2:29:24 PM
I am analyzing the genome data with univariate linear mixed models in gemma as a comparison to the BSLMM results (which are oddly uninformative given the amount of sequence data). I am first computing the centered kinship matrix. Next I am generating REML estimates of lambda (the ratio of variance explained by the kinship matrix to residual variance), and then testing individual SNP effects. Note, I modified the search region for lambda to avoid potential numerical under or overflow (I am worried this might explain the exceptionally low estimates of lambda). I am searching from 0.001 to 1000 with 20 (rather than 10) regions.
Here is the code:
cd /home/A01963476/projects/timema_wgexperiment/gemma/
gemma -g geno_timemaA.txt -p pheno_samplesA.txt -gk 1 -o timemaA
gemma -g geno_timemaA.txt -p pheno_samplesA.txt -a annotation.txt -k timemaA.cXX.txt -lmm 4 -o lmmout_timemaA -lmin 0.001 -lmax 1000 -region 20
cd /home/A01963476/projects/timema_wgexperiment/gemma/
gemma -g geno_timemaC.txt -p pheno_samplesC.txt -gk 1 -o timemaC
gemma -g geno_timemaC.txt -p pheno_samplesC.txt -a annotation.txt -k timemaC.cXX.txt -lmm 4 -o lmmout_timemaC -lmin 0.001 -lmax 1000 -region 20