Post by mocksu on Sept 5, 2021 3:03:37 GMT
I am trying to do two sample Mendelian Randomization. As the first step, I need to find independent signals (SNPs) that are associated with a protein level. I have a full summary stat file for the pQTL of the protein. I used ""gcta64 --bfile <my_ref_in_plink_bfile_format> --cojo-file <my_full_smmary_stats_file> --cojo-wind 20 --cojo-slct --extract <my_pval<5e-8_summary_stats_file> --out <my_output>".
The problem is that 6-10 independent signals are expected while I keep getting 1 independent signal, regardless the --cojo-wind size (10MB, 1MB, 200kb, 20kb, 10kb) and --cojo-collinear (0.9, 0.99 tried) used.
Not very sure what the "--bfile" option is for. My guess is that it's for the LD reference panel? HGDP + 1000 Genomes V3 3942 samples (downloaded from storage.googleapis.com/gcp-public-data--gnomad/release/3.1/vcf/genomes/gnomad.genomes.v3.1.hgdp_1kg_subset.chrXX.vcf.bgz and converted to bfile using plink) is used for the "--bfile" option. Would it be suggested to partition the 3942 samples into different ethnic groups and use the specific ethnic group as the ref panel? Or it does not matter?
Any comments/suggestions would be highly appreciated.
Thanks in advance!