|H. erectus (fem.) reconstruction|
Peter J. Waddell, Jorge Ramos, and Xi Tan. Homo denisova, Correspondence Spectral Analysis, Finite Sites Reticulate Hierarchical Coalescent Models and the Ron Jeremy Hypothesis. Pre-publishing at arXiv (2012, freely accessible).
This article shows how to fit reticulate finite and infinite sites sequence spectra to aligned data from five modern human genomes (San, Yoruba, French, Han and Papuan) plus two archaic humans (Denisovan and Neanderthal), to better infer demographic parameters. These include interbreeding between distinct lineages. Major improvements in the fit of the sequence spectrum are made with successively more complicated models. Findings include some evidence of a male biased gene flow from the Denisova lineage to Papuan ancestors and possibly even more archaic gene flow. It is unclear if there is evidence for more than one Neanderthal interbreeding, as the evidence suggesting this largely disappears when a finite sites model is fitted.
The paper needs some style revision but is otherwise very interesting, even if it's largely an exercise of statistical analysis, often resulting somewhat arid.
(...) one of the most surprising features of the planner NeighborNet model is that it does not reverse the positions of Neanderthal and Denisova, so that Papuan could have a unique split with the Denisovan (as Reich et al. 2010 suggest the Papuan lineage received ~5% of its genes from that lineage). As we will see later, the apparent reason for this would seem to be that the distance from Denisova to Chimp is more strongly underestimated than that from Denisova to Papuan. The underestimation of the Denisova to Chimp distance could be due to Denisova harboring some very archaic alleles, or it could be sequencing error.
The decrease in frequency of the DP pattern on X, particularly when compared to the NP pattern (which is near autosomal average frequency on X) suggests the possibility of asymmetric gene flow in this introgression event. If so, it would seem that this might be most readily explained by greater survival and reproduction of the offspring of Denisova males impregnating the modern human female ancestors of Papuans rather than the other way around.
Note the high frequency of the DNP pattern, which may be due to the Denisovan relatives that mixed not being closely related to the Denisovan sampled.
It seems tempting then to think that a model of three independent out of Africa lineages, with three independent mixings with the same population of Neanderthals (plus the independent Denisovan mixing event), would fit markedly better than the present model.
This last bit I find hard to believe, notably because we know of no Neanderthals ever existing in East Asia. In addition it would be a very odd coincidence that all three arrived to the approximate same amount of Neandrthal admixture. It seems much more likely that these smaller differences have been fixated as the three populations diverged in the Greater Eurasian expansion, after the OoA initial migration, when they probably incorporated the Neanderthal ancestry. (My two cents anyhow).
In any case, their conclusions follow:
Overall, the fitting shows that a hierarchical structured coalescent model with at least two introgression events between archaic humans and out of Africa Moderns leads to a substantial increase in fit. Overall fit however, is still far far worse than could be expected. It seems that to improve the fit a number of factors may come into play. Firstly, there are too many private NH, NF and NP [Neanderthal-Han, -French and -Papuan] patterns. Secondly, the latter of these, NP, seems markedly less than the former two. Thirdly, there may be too many sequencing/alignment errors in the present data to confidently move towards refining so many parameters and the overall fit. The marked improvement in fit when a finite sites model is employed is consistent with this. One model that may do a better job of describing the data with fewer parameters is independent mixing of Neanderthal genes with Han and French, but to a nearly identical total degree. Also, lesser mixing of Neanderthal genes into Papuan, made up for by a larger proportion of archaic alleles in Papuans coming from the mixing with an archaic that is only slightly closer to Denisova than to Neanderthal. This would in turn suggest that the mixing with Neanderthals was not purely right out of Africa and it was not a single event. Instead, there may have been opportunity for European ancestors to pick up Neanderthal alleles, in the unknown part of Eurasia they existed in prior to moving into Europe, ditto and independently for the ancestors of the East Asians, while Papuan ancestors moved fairly rapidly through the zone of classical Neanderthals and picked up most of their archaic genes in the Indonesian region. The form of this ancestral population may have been about equally related to Neanderthals and Denisovans, but may also have had an appreciable proportion of even earlier (e.g., Homo erectus genes) in its genome. This last point comes up in a number of analyses including the resampled NeighborNet and the finite sites model, but confirmation is difficult as the rate of sequencing / assembly error could be having a similar effect.
For background in this blog and its antecessor Leherensuge (from oldest to most recent):
- Neanderthal gene flow found in modern humans (on Green 2010)
- Exploring the Neanderthal admixture episode (version 1)
- Denisova hominins, Neanderthals, Melanesians and so on... (on Reich 2010)
- Explaining 'Denisovan' and also 'Neanderthal' admixture: the simplest scenario (where I first hypothesize that H. erectus and also maybe a relative of Neanderthals like the Hathnora hominin, could be the actual culprits).
- Is X-DNA lineage Neanderthal?
- Denisovan admixture widespread beyond Wallace Line, non-existent elsewhere - but then:
- Minimal Denisovan admixture in SE Asians.