I'm fairly new to plink software and wanted to get some additional practice after doing several tutorials. I obtained the data from this paper (I'm not using this paper's methods) to do some QC with subsequent analysis because they mostly use SNPstats.
Some background on what I did leading up to PCA:
My QC steps involved:
Filtering for low MAF
Filtering for high missing values on the individual and SNP level
Excluding outliers in heterozygosity (based on F statistic)
Excluding controls that deviated from Hardy-Weinberg
Finding related individuals and excluding members from the pairs that had the lower call rate
After these steps, I had 657649 variants remaining, and no individuals had been filtered out from the original total (1401)
I then used a 1000 genomes dataset that I had already downloaded (unsure of which phase, but it contained 690 individuals) and merged it with my data (did similar QC steps on the 1000 genomes data, and also used the --flip flag to fix strand issues).
I did a rooted PCA analysis (with a LD threshold of 0.2, which included 112795 variants post-filter, but I also tried PCA without the filter with similar results) to look at population substructure and got this plot (in R):
Where the pink "OWN' labels in the plot are my data points. Does anyone have any idea why my data wouldn't be lining up with any of the ethnic groups from 1000 genomes? The data is said by the authors to be "pre-filtered," which I'm assuming means fairly homogenous, so I wasn't expecting them to be in different ethnic groups, but they should be lining up with one of them, right?