Comments on: Tightening the interval of the expected

By: kjmtchl

kjmtchl — Thu, 07 Jul 2011 09:04:05 +0000

The major conclusion from this paper prompts a big “Duh!” in response. Of course rarer variants are population-specific. They are rare because they are recent, and/or being selected against. Most will be neutral but of those that have any phenotypic effect, vastly more will be deleterious than advantageous. Rare variants are therefore far more likely to contribute to disease. This means that genome-wide association studies combining diverse populations and looking only at globally common variants are not sampling the variants of highest interest (explaining why they have not found much, at least for disorders with a strong effect on fitness, where predisposing variants are highly unlikely to ever become common). In contrast, some genome-wide association studies have been performed by segregating samples into population-specific bins and then summing hits anywhere in a gene across populations to give a gene-wide association value (allowing for different variants in different populations to be playing a part). As any rare variant will necessarily arise on the background of some common haplotype, analysing the common SNPs can detect a signal due to a linked rare variant. It may thus be possible to go back to GWAS datasets with this approach and re-mine them for interesting hits. On the other hand, it is becoming easier just to do whole-exome or whole-genome sequencing, which has the advantage of revealing all the variation and possibly pinpointing the real pathogenic mutations.

Time to sequence!

By: ohwilleke

ohwilleke — Thu, 07 Jul 2011 02:57:40 +0000

“They estimated that the last common ancestor of Asians and Africans in their data set was on the order of ~50,000 years before the present. This is absolutely unsurprising. As they note this is entirely consonant with the archeological record. What is fascinating is the confidence: 45 to 69 thousand years over the 95% interval. This immediately seemed congenially narrow to me, and they confirm this by reviewing earlier estimates with noisier data sets which had much larger intervals.”

It seems as if the inclusion of exclusion of outlier populations, like Tibetans, Ainu, aboriginal Australians, Papuans, Khoisan, Ket, etc. could easily skew this a great deal, particularly if the people mostly descended from later waves greatly outnumber the people from early waves.

I could easily see excluding critical parts of the global population from the sample, even if those populations account for only 1%-2% of the total, could have many tens of thousands of years of impact on the age range.

By: John Hawks

John Hawks — Thu, 07 Jul 2011 00:05:01 +0000

Passages in the discussion may be there to satisfy reviewers with different results.

By: gcochran

gcochran — Wed, 06 Jul 2011 23:12:23 +0000

The estimate of the mutation rate they used was 2.36 x 10-8 per generation. Recent direct measurements from family triads show about 1.1 x 10-8 per generation. I believe that fresher number, if correct, would materially change their estimates of population split dates.

By: Razib Khan

Razib Khan — Wed, 06 Jul 2011 19:25:00 +0000

well, if u like alfred remember the HDGP browser and the HapMap browser.

By: Nathan M

Nathan M — Wed, 06 Jul 2011 19:22:12 +0000

I knew that, and I've read a few of the papers on the subject. However, I did find a website I'd bookmarked earlier that might help: ALFRED-The ALlele FREquency Database

By: Razib Khan

Razib Khan — Wed, 06 Jul 2011 17:26:06 +0000

#1, you want to look for “ancestrally informative markers.” these are at higher frequency than these alleles, but they are good at differentiating between populations (high Fst). you’ll find lists in papers online. google scholar it.

By: Nathan M

Nathan M — Wed, 06 Jul 2011 17:20:35 +0000

So theoretically, if the rarer SNPs were identified by population, we could pare down genetic datasets to exclude more common SNPs. That should cut down the time required for ADMIXTURE runs, or some new software based on this finding. Is there any database of allele frequency by population for SNPs?