Logistic Regression And Effect Sizes In Gwas

Question

[snp_cook12]

Hello, I have a question, I know for GWAS, case/control studies use logistic regression, what does it mean when the effect size for a SNP is negative? I'm new to the field so I'd really appreciate it if someone clarified this for me. I've also read that a lot of phenotypes are inverse normally transformed from their raw scores. What does that mean? Thanks!

Answer 1

[Devon Ryan]

Assuming that you're talking about the regression coefficients, which are an effect size for logistic regression models, a negative value means the opposite of a positive value. Generally, this would mean that whatever SNP you're looking at is negatively predictive for (i.e. predictive against) the phenotype in question.

[brentp]

I think a simpler way to say it is that the SNP is protective. Meaning having the SNP is associated with a decrease in the probability of having disease.

[Devon Ryan]

I actually used the "protective" nomenclature at first and then changed it. In reality, I don't think it's proper to call a SNP itself protective unless you then show with other experiments that it's not just simply associated with other changes that are actually the ones conveying the protective effect.

Answer 2

[snp_cook12]

thank you! that makes sense.