Human population history from single human genomes (Li & Durbin 2011)

I will update this blog entry when I read the paper. In the meantime, see Nature News and New Scientist.

UPDATE I

From the supplementary material (p. 8):

The TMRCA estimated by the PSMC model is in the units of mutation per site. To rescale TMRCA in the units of years, we need to know the mutation rate per site per year, which can be estimated by using closely related species. Table S1 implies that in primates, the mutation rate is broadly around 10−9 per site per year, the rate we used in rescaling the PSMC estimate (we assumed a 2.5 × 10−8 mutation rate per site per generation and a 25-year generation time, which is translated to a 1.0 × 10−9 mutation rate per site per year).

However, recent direct measurement using whole genome sequences in pedigrees suggest that in the individuals examined the mutation rate per site per generation approaches 10−8 (Roach et al.,2010; 1000 Genomes Project Consortium, 2010), twice smaller than the rate we use. Nonetheless,what matters for population genetic based methods such as PSMC is the time average. A comparatively small fraction of higher mutation rates could change this average significantly. We therefore feel that although direct measurements are clearly valuable, there are not enough yet to change the mutation rates used in population genetic based analyses.

The mutation rate was an issue in another recent paper, which used a similar 2.36x10-8 rate as the one here, and not the much lower rate from a couple of 1000 Genomes family trios.

As I said in that post (and more recently), we clearly have a lot to learn about autosomal mutation rates yet, and hopefully we will both get a better estimate of the rate from more trios of the 1000 Genomes project, as well as establish possible population variation in that rate.

UPDATE II (Divergence of Europeans and East Asians):

From the supplementary material (p. 13):

On the other hand, several studies using nuclear DNA placed the East Asian-European divergence around 17–25kya (Keinan et al., 2007; Garrigan et al., 2007; Gutenkunst et al., 2009). Our PSMC estimate from the combined Venter and YH X chromosomes is also very recent (Figure S7d). This leads to the apparent inconsistency with the fossil evidence that anatomically modern human have spread across the continent by at least 40kya. One of the possible explanations is that during the Last Glacial Maximum at about 20kya, the non-African populations retreated southward (Forster, 2004), and gene flows may have occurred between the different populations again. Under this hypothesis, the recent gene flow between YRI.X and KOR.X would be reasonable, although autosomal data from more populations are needed to further confirm the existence of the recent gene flow.

Gravel et al. suggested that there may have been "ghost populations" intermediate between Europeans and East Asians that suppress their divergence times; the explanation of Li & Durbin is different, but of the same kind.

An easy reconciliation of the archaeological divergence times with the genetic evidence, would, of course, be immediately effected if the "slow" family-derived rate is adopted: this would double West/East Eurasian split time to about 40kya, but would also push back the split of West Africans from Eurasians to the dawn of anatomical modernity to more than 200kya, and, the African hunter-gatherers (not examined here) well into multiregional evolution time depths.

UPDATE III (Jul 14): (A chicken and egg problem)

The authors use a 2.5 × 10−8 mutation rate per site per generation and a 25-year generation time in the paper, citing Nachman and Crowell (2000).

Nachman and Crowell estimate this rate with a Chimpanze-Human divergence at 5 million years and an ancestral population size of 10,000. However, since their generation length is 20 years, their 5 million years become 6.25 million in 25-year generation terms; the authors of the current paper (Table S1) put the human-chimp divergence at 7 million years.

What is most interesting, is that the current paper estimates ancestral population sizes by fixing the mutation rate; whereas Nachman and Crowell (2000) estimated the mutation rate by making different assumptions about ancestral population size. For example, their rate of 2.5x10-8 assumes an ancestral population size of 10,000 whereas for an ancestral population size of 100,000, this becomes 1.5x10-8.
In other words, it's a chicken and egg problem: the mutation rate has been calibrated on assumptions about ancestral population size in the earlier paper; ancestral population size is estimated by using the mutation rate in the current one.

I really do think that the way forward is to get a better estimate of the mutation rate from actual parents and children, because I see no obvious way to go around the above-mentioned problem.


UPDATE IV (Jul 14): (Possible population structure)

From the paper:

All populations showed increased Ne between 60 and 200 kyr ago, about the time of origin of anatomically modern humans17. An alternative to an increase in actual population size during this time would be that there was population structure involving separation and admixture11,16 (Supplementary Fig 5).

In the supplement, the authors consider a split into two or three sub-populations at 250ky followed by admixture at 60ky. In such a scenario, the pattern of growth between 200ky and 60ky can be explained without any actual growth taking place: the apparent growth is due to the admixture event between different types of humans.

I would also add the difference between the apparent severity of the Eurasian bottleneck after 60ky (compared to Africans) may also be due to the continuation of admixture in Africa which keeps the apparent effective size high, whereas Eurasians now begin to move outside Africa, and no longer have the opportunity to mix with archaic Africans.

UPDATE V (Jul 14): It is extremely unfortunate that this type of research was not carried out on Native Americans, Native Australians, and African hunter-gatherers. All of these would provide useful insight:

1. Native Americans, because they would be somewhat immune to "late" gene flow with Africans that is hypothesized to have affected even East Asians
2. Native Australians of Papuans, because of their substantial hypothesized "Denisovan" admixture which ought to register as an episode of "higher effective population size" prior to the admixture event
3. African hunter-gatherers, because they, more than anyone else, would push the limits of inference to the past.
   

Nature (2011) doi:10.1038/nature10231

Inference of human population history from individual whole-genome sequences

Heng Li & Richard Durbin

The history of human population size is important for understanding human evolution. Various studies1, 2, 3, 4, 5 have found evidence for a founder event (bottleneck) in East Asian and European populations, associated with the human dispersal out-of-Africa event around 60 thousand years (kyr) ago. However, these studies have had to assume simplified demographic models with few parameters, and they do not provide a precise date for the start and stop times of the bottleneck. Here, with fewer assumptions on population size changes, we present a more detailed history of human population sizes between approximately ten thousand and a million years ago, using the pairwise sequentially Markovian coalescent model applied to the complete diploid genome sequences of a Chinese male (YH)6, a Korean male (SJK)7, three European individuals (J. C. Venter8, NA12891 and NA12878 (ref. 9)) and two Yoruba males (NA18507 (ref. 10) and NA19239). We infer that European and Chinese populations had very similar population-size histories before 10–20 kyr ago. Both populations experienced a severe bottleneck 10–60 kyr ago, whereas African populations experienced a milder bottleneck from which they recovered earlier. All three populations have an elevated effective population size between 60 and 250 kyr ago, possibly due to population substructure11. We also infer that the differentiation of genetically modern humans may have started as early as 100–120 kyr ago12, but considerable genetic exchanges may still have occurred until 20–40 kyr ago.

Link