what is pc1 pc2 pc3 pc4 in a phenotype file for GWAS and how is this calculated

Question

[iphoenix2100]

Hi all,

I just moved from a molecular background and find it bit difficult to understand few terminologies in GWAS analysis.

I have this phenotype file with famly id, phenotype_measurement, sex, age and four other columns (pc1...pc4)

May I know what these prinicipal components are and how they are calculated.

Additionally I have another phenotype file that has information on family_id, phenotype_measurement, and age - Is it possible to calculate PCs for this file as well. IF so where to start.

Sorry for being so amateur...

Thanks
chr

Answer 1

[Devon Ryan]

For how PCA is calculated have a look at the wikipedia page (note, you might need to have taken linear algebra to really understand things completely). In general, PCA on a dataset like yours is used to given an idea about how the population is stratified and the principal components can then be used as covariates in the GWAS calculations (so you don't get spurious findings due to samples being from different populations).

A lot of people use plink for things like this and you might have luck with it as well. It's a powerful tool, but I expect the learning curve isn't terribly shallow, so you might try to find someone local who's done GWAS experiments before to help you out the first time.

Answer 2

[BioCh'ti]

Personnaly, I always use the PCA components as phenotypes to check whether if the MLM model will still detect SNP associated to the population stratification. If not, it means that you do not have any bias introduced by the population stratification and do not overstimate p-values (visualize by qqplots). This is particularly true when you have phenotypes linked to population stratification, because, when you correct for it, by introducing your Q matrix in the MLM model, you also 'hide' interesting SNPs.