GWAS on unbalanced case-control UKB dataset using PLINK2

Question

[DobraVila]

Hi everyone,

I want to run GWAS of Schizophrenia. However, the problem is that the data is unbalanced. There are 606 cases and 133,000 controls.

1. What would be the best way to perform GWAS? Is it even worth doing it given the fact that there are only 606 affected individuals?
2. Can I use PLINK2 to run GWAS on an unbalanced dataset? I noticed that some studies recommend SAIGE for unbalanced datasets.

Thanks!

Answer 1

[bk11]

What would be the best way to perform GWAS? Is it even worth doing it given the fact that there are only 606 affected individuals?

Yes, it is still worth performing GWAS with 606 cases. If you want to perform GWAS using balanced dataset, you can perform PCA match with appropriate case:control ratio using PCAmatchR tool.

Can I use PLINK2 to run GWAS on an unbalanced dataset? I noticed that some studies recommend SAIGE for unbalanced datasets.

Yes, you can use PLINK2 too to run GWAS on an unbalanced dataset. PLINK2 has the Firth regression model added to improve handling of rare variants and imbalanced binary phenotypes. SAGIE also handles the imbalanced binary phenotypes but it relatively slow.