population expansion

Jay F, Boitard S & Austerlitz F 2019 An ABC method for whole-genome sequence data: inferring paleolithic and neolithic human expansions. Mol Biol Evol 36:1565-1579.

  • [based on the preprint version downloaded at bioRχiv]
  • the inferred effective sizes and timings of expansions vary widely across studies and seem sensitive to the inference method and sample size used
  • some studies inferred an expansion starting in the upper Paleolithic
  • other studies using rapidly mutating microsatellite markers (Aimé et al. 2014) or specific mitochondrial haplogroups (Soares et al. 2012) point toward a more recent Neolithic expansion which is consistent with the hypothesis of two subsequent expansions (Tennessen et al. 2012)
  • these studies provide thus an indirect evidence of this two-expansion process
  • the question remains whether we can infer this process directly from whole-genome data
  • detecting two subsequent expansions
  • the probability of identifying the model One expansion while in fact Two expansions occurred increased when the ancient and recent growth rates were very similar and when the old expansion was too weak compared to the recent expansion
  • two expansions were identified when the ancient growth rate was larger than 1×10−3, or between ~3.5×10−4 and 1×10−3 if the recent rate was larger
  • the bottleneck strength had an additional impact on misclassification
  • datasets generated under a Bottleneck and two expansions scenario were more commonly assigned to Bottleneck and one expansion if the bottleneck was strong and recent
  • demography in European, South and East Asian groups
  • we first performed model selection on all populations and systematically identified the bottleneck followed by two expansions as the most likely scenario for non-African populations
  • posterior probabilities associated with this model ranged from 0.78 to 1 (table 3)
  • we found evidence for a strong bottleneck, with an estimated reduction in size ranging between 0.0027 and 0.0126 (average 0.006), starting around 1869 generations ago on average across all populations
  • the bottleneck was followed by an ancient mild expansion with a population size reaching on average 84,205 (average growth rate = 0.003), and by a very recent and strong expansion that led to a current effective population size of 460,392 (ave. growth rate = 0.031)
  • the drastic change in growth rate occurred around 106 generations ago (fig. 4 and table 3)
  • we also inferred the parameters using the smaller tolerance rate of 0.001, which was found to give similar prediction errors but less accurate bounds for the 90% credible intervals
  • in that case, the estimated timing of the recent expansion and the bottleneck were slightly older (166 and 2322 gen. ago respectively, while the estimated current population size was lower (105,239 individuals)
  • even if the point estimates for the current population size differ to some extent, a recent strong expansion was found in both cases (recent growth rate of 0.020 versus ancient rate of 0.003)
  • demographic inference for African groups
  • when analysing each African population independently, we found that the model selected via ABC for each of these populations was the model assuming a bottleneck and a single expansion (probabilities 0.55-0.99)
  • for all five populations, we found that the bottleneck was very recent and strong (409 generations ago on average, effective population size reduced by a factor 0.0095 on average), and was followed by a drastic expansion (average exponential growth rate 0.036)
  • European and Asian histories
  • bottleneck and two successive expansions
  • for all populations pf European and Asian ancestry we detected the Out-Of-Africa bottleneck followed by two successive expansions
  • the ancient expansion was estimated to have started on average 1,190 [?] generations ago
  • ie 35,700 ya assuming a generation length of 30 years
  • the recent expansion 106 generations ago
  • i.e 3,180 years ago
  • the most ancient expansion appears thus as a signal of a Paleolithic expansion
  • the more recent expansion is consistent with the Neolithic transitions that emerged from 11,500 to 3,500 years ago across the world
  • this is of interest as only few studies had been able to detect simultaneously both expansions
  • demographic histories in Africa and model-based approach
  • the simple demographic scenarios, still often used to depict African populations, were not able to explain the observed data
  • we estimated the parameters of the Bottleneck and one expansion scenario based on classical statistics (heterozygosity, Tajima’s D, …) and the site-frequency-spectrum
  • we identified a weak bottleneck (average coefficient 0.75) around 3714 generations ago followed by a mild expansion (average growth rate = 4.4×10−4)
  • this scenario is not favored when exploiting numerous linkage-informed statistics
  • future developments are needed to address the challenging task of co-estimating selection and demography