Hi-
I ran the GWAS analysis using SAIGE and I had a very nosiy QQ plot and Manhattan plot. I just realized that I have not run QC steps on the imputed data. Is there a good resource on the QC steps before SAIGE analyses?
Best-
Nihal
Thank you Phil so much!I saw this very helpful thread.
I am thinking that I need the QC steps to remove individuals with missingness, relatedness, sex-genetic mismatch. And also remove these variants with low MAC. Are any new QC steps needed?
missingness and minor allele frequency/count is in the github files. You accomplish this using plink much the way saige uses plink for ld-pruning. See here for WES.
Comments
4 comments
Nihal,
Saige, plink, and regenie all require similar QC steps prior to GWAS.
Depending on the datasource: base-GT, Imputed-GT, or WES, You will have very slightly different QC parameters.
I have a repo on github with qc steps using swiss-army-knife for plink and regenie, and for the original imputed dataset as well as the WES data.
https://github.com/pjgreer/ukb-rap-tools/tree/main/GWAS_pipeline
much of that work was based on anaztazia's work here:
https://github.com/dnanexus/UKB_RAP
I don't recommend filtering on HWE and hope to have a paper on why soon.
-Phil
Thank you Phil so much!I saw this very helpful thread.
I am thinking that I need the QC steps to remove individuals with missingness, relatedness, sex-genetic mismatch. And also remove these variants with low MAC. Are any new QC steps needed?
Best-
Nihal
Removing highly related individuals and sex mismatch can be done when building the phenotype file.
There is a tutorial using python here, but I prefer using R.
https://dnanexus.gitbook.io/uk-biobank-rap/science-corner/gwas-ex
missingness and minor allele frequency/count is in the github files. You accomplish this using plink much the way saige uses plink for ld-pruning. See here for WES.
https://github.com/pjgreer/ukb-rap-tools/blob/main/GWAS_pipeline/gwas_wes38_plink/11b-gwas-s2-wes38-qc-filter.sh
Thank you Phil!
Please sign in to leave a comment.