Our estimate of the average common ancestor time reflects the average time at which the Neanderthal and human reference sequences began to diverge in the common ancestral population, not the actual split time of the ancestral populations that gave rise to Neanderthals and modern humans. To estimate the actual split time of the ancestral human and Neanderthal populations, we developed a method that incorporated data from the human and Neanderthal reference sequences, as well as genotypes from 210 individuals with genome-wide single-nucleotide polymorphism (SNP) data collected by the International HapMap Consortium () (19
). We included the HapMap data because they indicate what proportion of sites in the Neanderthal sequence fall within the spectrum of modern human variation. For example, if the ancestral human and Neanderthal populations split long ago, before the rise of most modern human genetic diversity captured by the HapMap data, then Neanderthal sequence would almost never carry the derived allele, relative to the orthologous chimpanzee sequence, for a human SNP (). Conversely, a more recent population split would result in Neanderthal sequence frequently carrying the derived allele for human SNPs.
Table 2 Summary of all autosomal sites sequenced in Neanderthal and uniquely aligned to the human and chimpanzee reference sequences. The designations “ancestral” and “derived” indicate whether each site is, respectively, a match (more ...)
To formalize this idea, we considered an explicit population model for the relationship between Neanderthals and each HapMap population (East Asians, Europeans, and Yoruba) separately (fig. S3) (13
). We assumed that Neanderthals and modern humans evolved from a single ancestral population of 10,000 individuals and that the Neanderthal population split away from the human ancestral population instantaneously at a time T
in the past, with no subsequent gene flow. In order to model the demographic histories of the HapMap populations, we made use of models and parameters estimated by Voight et al
) based on resequencing data from 50 unlinked, noncoding regions. Those demographic models include bottlenecks for East Asians and Europeans and modest exponential growth for Yoruba (13
We then constructed a simulation-based composite likelihood framework to estimate the time of the human-Neanderthal population split (13
). At each site in the human-Neanderthal-chimpanzee alignments we constructed, we recorded the Neanderthal and human reference alleles relative to chimpanzee. We also determined, separately for each population, whether each site was a HapMap SNP in that population and if so, the allele frequency (). We used simulations to estimate the probability of each possible data configuration at a single site as a function of the human-Neanderthal split time. The simulations used the estimated population demography for each HapMap population and a probabilistic model of SNP ascertainment to match the overall density and frequency spectrum of HapMap Phase II SNPs. Likelihood curves for the split time were computed by multiplying likelihoods across sites as though they were independent. In practice, this is an excellent approximation for our data because the Neanderthal sequence reads are very short and just 1 out of 905 aligned fragments contains more than one human-specific allele or SNP. Bootstrap simulations confirmed that our composite likelihood method yields appropriate CIs for the split time (13
Using this approach, the maximum likelihood estimates for the split time of the ancestral human and Neanderthal populations are 440,000 years (95% CI of 170,000 to 620,000 years) based on the European data, 390,000 years (170,000 to 670,000 years) for East Asians, and 290,000 years (120,000 to 570,000 years) for Yoruba ( and ). These values predate the earliest known appearance of anatomically modern humans in Africa ~195,000 years ago (22
). Because these split times are before the migration of modern humans out of Africa, the three population-specific estimates should all be estimates of the same actual split time. The average of these estimates, ~370,000 years, is thus a sensible point estimate for the split time. Substantial contamination with modern human DNA would cause these estimates to be artificially low, but 2% contamination, the rate suggested by mitochondrial PCR analysis of the primary extract used to construct the library, would have essentially no impact (13
Our estimates of the human-Neanderthal split time might depend heavily on the assumption that the ancestral effective population size of humans was 10,000 individuals. To address this, we explored a set of models in which the ancestral human population expanded or contracted at least 200,000 years ago (13
). We found that much of the parameter space—though not the original model—could be excluded on the basis of modern human polymorphism data from Voight et al
). We repeated our likelihood analysis of the Neanderthal data using models incorporating ancient expansion or contraction that are consistent with modern data and found that these did not substantially change our population split time estimates ().
Our data include three sites at which Neanderthal carries the derived allele for a polymorphic HapMap SNP. These sites are unlikely to represent modern contamination because for two of the SNPs, the derived allele is found only in Yoruba; also, one of the SNPs lies on a fragment that contains a C-to-T transition in Neanderthals that is characteristic of chemical damage to DNA. These observations indicate that the Neanderthal sequence may often coalesce within the human ancestral tree. Based on simulations of our best-fit model for Yoruba, we estimate that Neanderthal is a true outgroup for approximately 14% (assuming a split time of 290,000 years, the Yoruba estimate) to 26% (assuming a split time of 440,000 years, the European estimate) of the autosomal genome of modern humans, although more data will be required to achieve a precise estimate.