Test for genetic additive model using R

Question

[brunobsouzaa]

Hi everyone. I'm studying two SNPs in a case/control study and one part of the analysis is to test for additive model. So the question is: Can I use logistic regression for this task or should I just build a 2x2 table and perform Fisher exact test? The data looks like this:

Outcome SNP_1 SNP_2 Gender Age
1       AA    GG    M      45
0       AC    GT    F      54
1       CC    TT    F      65

Outcome = 0 (Control); 1 (Case)

SNP_1: A = Risk

SNP_2: T = Risk

For logistic Regression, how can I format my data to fit in this kind of model? Should I just convert my genotypes to 0, 1, 2, merge both SNPs into a single column and use Outcome ~ SNP, family = binomial or there's another way to do this?

Thanks.

Answer 1

[rpolicastro]

It may be worth testing an interaction term as well, such as Outcome ~ SNP_1 + SNP_2 + SNP_1:SNP_2, or the R shortcut Outcome ~ SNP_1 * SNP_2. This model would not only take into account the main effects of SNP_1 and SNP_2, but also test whether any SNP_2 SNPs influence the effect of any SNP_1 SNPs. The effectiveness of this model will somewhat depend on the sample size, since you want a good number of samples for each factor level.

[brunobsouzaa]

Thank you very much, man! Upvoted!