Can someone please help me understand how exactly I should be interpreting a PCA plot?
I have read up at length on the maths behind it but there doesn't seem to be much reading available on how one actually goes about interpreting them as far as I have been able to find?
Some Background:
I am performing research on SNPs and population genetics using PLINK and (very nicely annotated) VCF's but with no case/control, just SNP's. I have seen in PLINK's documentation that the package is able to generate variance-standardised relationship matrix based PCA's. I have successfully managed to generate it using just a simple plink --bfile <prefix> --pca header --out <outPrefix> which is working hunky dorey, spitting out the top 20 Eigenvalues. It became apparent however, that as soon as I graphed it using R, that I'm not entirely sure whats happening or how to read the graph. I know the principle components themselves are meaningless (I say this tentatively), but then what exactly should this graph be telling me about my data? I know from previous contextual usages I have seen in presentations that I am supposed to be looking for clusters and groups but now what do these clusters actually mean in a biological context?
Ok, So I have been reading up more on PCA's and asking around and from what I have gathered, the gold in a PCA is being able to overlay the plotted points with cluster information to see which sets of PC's cause the most migration on the data (I.e. explain the structure of the data the most). From what I have been told, ideally, you would then link that back to the exact (in my case) SNP that was used to generate that PC.
So I guess now my question becomes, is there any way in PLINK to trace the PC's back to SNP's? I have not seen anything to that effect in their documentation.
That may very well be a question for the PLINK developer, who can take up the question at next login.