Assessing Ancient DNA Sampling Strategies for Natural Selection Inference in Humans Using Allele Frequency Time...

Area of Science:

• Population Genetics
• Paleogenomics
• Evolutionary Biology

Background:

• Genomic data from ancient humans enable natural selection estimation using time series.
• Existing methods were validated with large sample sizes, not sparse ancient DNA (aDNA) data.
• aDNA data present challenges like high missingness and uneven sample sizes per time point.

Purpose of the Study:

• Benchmark selection inference methods using aDNA-like time series datasets.
• Evaluate method performance with sparse data and small sample sizes.
• Determine feasibility of detecting selection in ancient human populations.

Main Methods:

• Extensive simulations using aDNA-like time series datasets.
• Benchmarking four methods: ApproxWF, BMWS, Slattice, and Sr.
• Testing various sampling schemes and selection coefficients (s).

Main Results:

• Selection detection is possible for strong selection (s ≥ 0.02) with ~100 samples, assuming constant effective population size (Ne).
• ApproxWF demonstrated superior performance across simulations compared to other methods.
• Low false positive rates (<6%) and high power (>90%) for detecting strong selection were observed.
• Homogenous data sampling across time points improved estimation accuracy.

Conclusions:

• Inferring natural selection from sparse aDNA data is achievable, particularly for strong selection.
• ApproxWF is recommended for analyzing aDNA time series data.
• Future research should focus on even data distribution across time for improved accuracy.