Filtering GWAS output by pvalue prior to submission

[Daniel Rodriguez]

Hello,

I submitted by GWAS output and got some ok results. But when I submitted my GWAS output only including SNPs that reached p<0.05 in my GWAS, the genes and pathways were a lot more in line with the clinical outcome measured in my GWAS.

I understand that the pvalues from MAGMA will be artificially inflated but is there a way to still use these results? Is there a reason why you still include very insignificant snps in the analysis?

Thanks so much!

[Tanya Phung]

Dear Daniel, 

Thank you for using FUMA and for your message. 

MAGMA is built based on a model with the p-values from all of the variants included. Therefore you are right that when you are selecting a subset of variants based on significance of pvalues, it's going to bias the model, and the results might not be reliable. 

I would suggest that you consult the manual of MAGMA (https://ctg.cncr.nl/software/magma) and if you still have additional question, feel free to send a message to the developer of MAGMA. 

Thank you,
Tanya