Post date: Jan 20, 2015 8:20:18 PM
Given the mostly underwhelming results from analyzing the whole genome data and our inability to perform cross-validation with these data, I am now looking back at the GBS data. The idea is to fit the GBS data in gemma, see what we get for PVE and no. SNPs, check the model with cross-validation and test for tradeoffs at the SNP level. We can then follow-up with scans of specific regions of interest with the whole genome data as needed.
I would like to re-assemble and re-call variants with the GBS data, but I am waiting on a barcode file from Tom/Victor (these data were assembled and called prior to the switch to bwa and genotype likelihoods with error). So, I am working with the old calls and genotype likelihoods for now just to see if there is any signal. Everything is in /labs/evolution/projects/timema_wgexperiment/gbsgemma/. For now I am running 10 chains with 500,000 burnin, 1.5 million post burnin steps, and a thinning intervals of 20, e.g. command:
cd /home/A01963476/projects/timema_wgexperiment/igbsgemma/
sleep 45
gemma -g geno_gbstimemaC.txt -p pheno_gbssamplesC.txt -bslmm 3 -n 1 -o out_gbsgemma_gbstimemaCrep9 -rpace 20 -maf 0.01 -w 500000 -s 1500000