Considerations when applying FOCUS to GWAS hits with p-value less than 5e-08

[NaomiHuntley]

Thank you for creating this tool!

I have run the FOCUS analysis on my dataset but I am deciding if it is best to relax the p-value or keep it at 5e-08.

For many of the traits I am looking at only 1 - 8 SNPs that reach the genome wide significance cut off. Can you advise on if there is a minimum number of SNPs that is necessary to run this analysis? I am trying to understand if it is 'better' to run FOCUS on a few SNPs that are very significant, or to relax the p-value so that I can include additional suggestive SNPs.

Thank you in advance!
~ Naomi

[zeyunlu]

Hi @NaomiHuntley,

I hope this message finds you well. Thank you for trying our software and for your thoughtful question!

This is a great question. FOCUS is a TWAS fine-mapping method based on gene-level TWAS statistics. In other words, it operates independently of whether your GWAS has genome-wide significant signals. Methods like FUSION compute TWAS statistics by first fitting an expression prediction model without considering GWAS signals. They then impute TWAS statistics using GWAS summary statistics without factoring in genome-wide significance.

That said, as a best practice, we typically recommend performing TWAS fine-mapping in regions (i.e., LD blocks) that contain GWAS signals. If you have eight GWAS-significant regions, FOCUS will analyze each of them. If you find the standard threshold of 5e-8 too strict, you can relax it to include more regions in fine-mapping. The decision ultimately depends on your judgment and prior knowledge of the trait you’re studying.

Please feel free to reach out if you have any further questions!

Best,
Zeyun