NEW YORK (GenomeWeb) – With a new method to infer population size changes from SNP data, a research duo from the University of Texas Health Science Center has examined fluctuations in human population size dating back tens of thousands of years.
As they described in Nature Genetics today, Texas' Xiaoming Liu and Yun-Xin Fu developed a model-flexible method called a stairway plot that uses SNP frequency spectra data to gauge population size. Liu and Fu then applied this approach to study nine populations from the 1000 Genomes Project to find evidence of a population bottleneck in European populations, though not in Finns, between 20,000 years and 30,000 years ago and a bottleneck some 100,000 years to 200,000 years ago in African populations.
Many approaches for inferring demographic history, the pair said, rely on predefined population models and are thus better suited for hypothesis testing rather than for exploratory or hypothesis-generating analysis. At the same time, model-flexible plots like skyline plots can only be applied to recombination-free loci, and those like pairwise sequentially Markovian coalescent framework-based models require a broad idea of population history as well as high-quality sequence data and can produce estimates that are biased for recent population history.
Their stairway method instead uses a flexible, multi-epoch model, like that of a skyline plot, that calculates the expected composite likelihood of a given SNP frequency spectrum. This, they noted, treats each SNP as an independent locus, a move that reduces the computational burden of the approach.
"Therefore, the stairway plot has both the model flexibility of the skyline plot methods and the computational efficiency, making it applicable to hundreds of individuals," Liu and Fu said in their paper.
Validation of their approach using simulations indicated that it provided estimations of recent population size changes that were more accurate than Markovian coalescent framework-based models.
The pair then applied their method to whole-genome sequences from nine 1000 Genomes populations — Utah residents with Northern and Western European ancestry; British in England and Scotland; Toscani in Italy; Finns; Han Chinese in Beijing; Southern Han Chinese; Japanese in Tokyo; Yoruba in Nigeria; and Luhya in Kenya.
The researchers restricted their analyses to regions located more than 50 kilobases from any coding region, those within the strict mask of the 1000 Genomes Project, and those whose ancestral alleles had been inferred with high confidence.
For each of the nine populations, Liu and Fu created 200 bootstrap SNP frequency spectra (SFS) and for each bootstrap SFS, the stairway plot was used to infer demographic history. The median inferred population in each time interval from each of the 200 estimations was used to develop a single inferred history of population size.
All the non-African populations, the researchers reported, exhibited severe population bottlenecks between 50,000 years and 70,000 years ago, likely corresponding to the out-of-Africa migration.
In addition, all the non-African populations, except for the Finns, had another shallower and more recent bottleneck that occurred some 20,000 years to 30,000 years ago and was followed by a recovery in population size.
The ancestors of Finns, instead, appeared to have had a population bottleneck between 10,000 years and 20,000 years ago that was not seen in other European populations. One possible explanation for this, the researchers said, is that the Finnish ancestors separated from other Europeans some 30,000 years ago or that they experienced the same bottleneck 20,000 years to 30,000 years ago that other Europeans did, but then experienced another one some 10,000 years to 20,000 years ago.
According to additional simulations the researchers ran, their approach wouldn't be able to infer both bottlenecks in that scenario and would likely pull out only the more recent one, as it appears to have done.
Meanwhile, the two African populations experienced a wider and shallower population bottleneck some 50,000 years and 70,000 years ago, and no bottleneck between 20,000 years and 30,000 years ago.
In addition, though, the African populations harbored evidence of a bottleneck tracing back some 100,000 years to 200,000 years ago that was not seen in the non-African populations.
Similar to the Finnish group, the researchers said that it's possible that the ancestors of all non-Africans separated from the ancestors of Africans 200,000 years ago, or that their approach doesn't have sufficient power to infer this ancient bottleneck in non-Africans.
"These results also emphasize that interpretations of inferred bottlenecks should be undertaken with care and that hypothesis testing is necessary before any conclusions are formulated," the pair noted.
They suggested that their approach be used as a complement to other methods to infer demographic history, and noted some tweaks that could be made such as developing a more efficient optimization search algorithm to reduce the computational load and enable the use of larger samples.