is it completely necessary to include principal components in the analysis??

[corunesa]

Hello! I am writing because I wanted to know if it is completely necessary to include principal components in the regression. I understood that saige uses LMM and reading about this kind of models, I understood that it was controversial to add them, although in the SAIGE article they do. Also, in my case, PC1 explains 1-4% of the variance, and is not associated with my phenotype of interest in a multivariate regression with non-genetic data. Could someone help me? Thank you very much!

Translated with www.DeepL.com/Translator (free version)

[rkarlssonlinner]

Hi,

It is recommended to include genetic PCs also when estimating LMM because the exogeneity assumption of the random effects model can be violated, and because population stratification can be considered a fixed effect rather than a random effect. See e.g., Price et al. (2010): https://www.nature.com/articles/nrg2813

"However, population structure is actually a fixed effect (that is, its effect as a function of genetic ancestry is the same for all samples), and spurious associations might result if it is modelled as a random effect based on overall covariance, particularly in the case of unusually differentiated markers. Modelling population structure as a fixed effect provides a higher level of certainty in correcting for stratification but requires running PCA (or a similar method) to infer the genetic ancestry of each sample."

[corunesa]

Thank you very much, I understand :)

[Mahmedturk]

do we add the PCs as co-variates in the phenotype file?

[rkarlssonlinner]

"do we add the PCs as co-variates in the phenotype file?" That is how I typically do it.