Principal Component Analysis using SNP data ste

Question

[SOHAIL]

Hi everybody,

I have two population of Whole genome SNP calls, and i want to perform Principal component analysis on my set of variant calls together with variant information of population samples from 1000 genomes by using Eigensoft or some other good software.

can anyone please describe in steps how to do that, especially (bit in detail) how to combine the variant information between 1000 genome samples and my samples and later format conversion steps?

Thank you very much! .

Answer 1

[Ahill]

If you are an R user, the SNPRelate package provides PCA and routines that would allow you to bring in datasets like 1000G from VCF or PLINK format files: http://corearray.sourceforge.net/tutorials/SNPRelate/#principal-component-analysis-pca http://dx.doi.org/10.1093/bioinformatics/bts606

[SOHAIL]

Hi Ahil! cannot access openthens account.. :(

Answer 2

[WouterDeCoster]

I never got to play with it, but kept it saved in my to do list (probably forever): http://alimanfoo.github.io/2015/09/28/fast-pca.html

Answer 3

[brentp]

If your data is in VCF format, you can do this with peddy: https://github.com/brentp/peddy

The command would look like:

python -m peddy -p 4 --plot --prefix my.out $vcf $ped

this will do a PCA with your samples projected onto those from thousand genomes and ancestry predicted according to those. In addition to an interactive html file, you'll get (among others) a PCA plot that looks like:

Where the points from your cohort will be the big open circles and the small ones in the background are the 1000 genomes samples.