Up-to-date GWAS Power Calculator and GWAS power comparison

Question

[moxu]

I have used two different methods to do a height GWAS, and would like to know which method has more statistical power. I've searched over the internet and found that most software products are either

• not available anymore (obsolete links) or
• only available online (implying manual input for millions of snps) or
• irrelevant (e.g. to help calculate sample size for a to-be-planned GWAS).

The only package seems to be appropriate is GWAS power at https://github.com/kaustubhad/gwas-power, but I have a question regarding which p-value to use (https://github.com/kaustubhad/gwas-power/issues/3) and not sure when the question will be answered.

The above being said, I am wondering how people are calculating GWAS power these days.

A related question is what metrics to use to tell which GWAS method is better than the other besides power. Suggestions would be appreciated.

Thanks much in advance!

Answer 1

[Kevin Blighe]

Hi,

Please try CaTS: http://csg.sph.umich.edu/abecasis/cats/

Kevin

[moxu]

Hi Kevin,

Thanks for the reply. I had checked out CaTS before posting above. However, I am not sure if CaTS does what I want to do.

Suppose I have two linear regression models for GWAS:

Model 1: height = age + gender + PC1 + ... + PC20 + SNP
Model 2: height = age + gender + SNP

After the GWAS is done, that is to say, all the betas, SEs, p-values are generated for each of the models, I want to see which model gives more power. Can CaTS calculate the actual power of the models so that we can compare which model is more powerful? If I understand correctly, CaTS only estimates (not calculates?) the power of a GWAS design before the GWAS is conducted.

Thanks.