Plink: Why No Cluster In Ibs Clustering For Multidimensional Scaling Plots?

Question

[w0nderingatwork]

Dear all,

is anyone firm in plinks '--cluster --mds-plot n' -function to calculate principal components for GWAS? Actually, I do get principal components when applying this function to my GWAS (3000 samples, 200k SNPs) dataset, but all my samples are within the same cluster ('cluster 0'). It is the same when I use just the '--cluster' function alone. I wonder if something's gone wrong, but there is no error message in the log file. Thanks for your help, best regards, Frank

Answer 1

[chrchang523]

To force a minimum number of clusters in the MDS plot, use the --K flag (e.g. "--K 3" for three clusters).

Meanwhile, note that PLINK MDS dimensions aren't the same thing as principal components (though they are similar enough that the difference probably won't kill you). GCTA and PLINK 1.9 have a simple --pca command to calculate actual principal components; if that's not flexible enough for your needs, try EIGENSOFT.

[w0nderingatwork]

thanks a lot. Actually, there is no certain number of clusters I want (just need the PCs). I am just wondering if anything went wrong because no clusters have been generated and therefore the PCs may have been calculated wrongly, too? But your advice to use the --pca option in plink 1.9 is very helpful, I am doing it that way now. Thank you!