Dear L. T. ,

Yes, I think that can work, although I would have thought Compound A concentration would have been a fixed factor. Make sure the assumptions are satisfied.

This is not a problem you can address with ANOVA.

You might consider an ANCOVA model on X with genotype groups and Compound A concentration as the cofactor. You would need one model for each other compound, which might be tedious. A MANOVA model controlling for "A" might do all compounds at once for you, depending on your data and the software you have available.

I think a software of SPSS can help you address this issue. this is a performing link: text introduction (https://www.youtube.com/watch?v=cOmbrK-TOqo) or video performing(https://www.google.com.hk/search?safe=strict&hl=zh-CN&ei=gyYXXZO2HZOYr7wPx6qVuA4&q=two+way+anova+spss&oq=+twoANOVA+SPSS&gs_l=psy-ab.1.0.0i7i30i19l10.16737.20239..22296...1.0..0.194.759.0j5......0....1..gws-wiz.......0i67j0i7i30j0i10j0i13.Z3y86wXKdoY)

If a higher concentration of compound A forces a lower concentration of other compounds, then thigh might need to be corrected for. In an experimental design study where you can manually vary the concentrations this is a mixture design but for your situation since you cant manually vary the concentrations idk maybe just a giant correlation study or ridge regression or lasso . We need more information about your study http://reliawiki.org/index.php/Mixture_Design

Yes it should be some kind of correlation but I have to take the genotype into consideration. So I jsut want to know woth which compounds the compound A correlates. There is not mcuh else to tell. I have a huge dataset and the effects of compound A + genotype.

I am working with R.