So I'm trying to analyze population genetic structure using a large data set of SNPs (~150,000 of them). I have 70 individuals sampled from 11 populations, and I am trying to assign them to either 7 or 8 clusters - as determined by comparing runs for various values of k. However, whenever I run STRUCTURE with k=7 or 8, the results always include one cluster to which no individuals are assigned. The Q table reports a column of zeros for that cluster, and not one individual is assigned a non-zero probability for membership in that cluster. The weird thing is that the unused cluster is still given non-zero values for expected heterozygosity and mean Fst.
Can anybody think of what might be going on here? I'm using only biallelic SNPs, with very little missing data, minor allele frequencies above ~16%, minimum depth of 6.
Thanks!