Low percentage explained by PCs

Question

[Tam]

Hi everyone,

I'm trying to do PCA on a VCF dataset with about 200 samples and 800k SNPs using the SNPRelate library. However, the variance explained by the first component seems to be very low (around 0.5%), as opposed to the example given by the tutorial (which as I understand has far fewer SNPs). I just want to ask if this is to be expected, if anyone has done this kind of analysis before.

Thank you

Answer 1

[Kevin Blighe]

If it's a homogenous population, then, yes, this would be expected. Are the samples from the same geographical region?

[Tam]

Thanks Kevin. The samples are indeed from the same geographical region. By homogenous, I assume you mean, e.g. having similar ethnicities, right?

[Kevin Blighe]

Yes, homogenous in that sense, i.e., a non-diverse ethnic pool. There may also be other reasons pertaining to any filtering that you have performed for, e.g., minor allele frequency or linkage disequilibrium.