The question: Is there currently any way to get a larger number of whole genomes (in a subpopulation specific fashion) than is available through CGI or 1kG? If I could obtain, say 1000 YRI, 1000 CEU, and 1000 EUR genomes, I think I could get accurate enough estimates for my intended goals, but really anything would help. I am comfortable with any format; whether web tool, vcf file, etc.
Background: Periodically, people ask if there are additional sources of publicly available NGS data in addition to 1kG and CGI data. Here is one such thread here: Sources Of Publicly Available Human Whole Genome Sequence Data
However, my question has two specifics, and in addition I want to post it in case something has become available recently. The specifics are:
1) The cohort would ideally have a moderate number of subjects from multiple global populations 2) Must be whole genome sequencing, not exome only.
So, from these posts, it is evident people have suggested TCGA, Kaviar, etc., but for the purpose of obtaining this would be to get accurate estimates of MAF of specific variants in a population specific way, I am not sure if these are ideal.
Of course, this is currently possible through a variety of sources; e.g. VEP or directly from the 1kg website; both of these draw on the same data, ultimately.
Here is the problem: Suppose that I am interested specifically only in Yorubans. Using any of these methods, if I go select a particualr variant, then trim the .vcf to the relevant sample_IDs only, I am going to end up with approximately 108 subjects (some of whom are related) and therefore about 216 total observations.
The expected variance for the allele frequency for this number of observations will in many cases be fairly high, especially for lower MAF.
Perhaps not exactly what you are looking for (no individual level data) but I saw no mention of gnomad: http://gnomad.broadinstitute.org
Huh. Nice. Was not aware of this resource.
This is actually very close to what I need. If it had African populations broken down more, I would be good to go. Thank you very much for passing along.