Non-parametric estimation of population size changes from the site frequency spectrum

Area of Science:

• Population genetics
• Evolutionary biology
• Bioinformatics

Background:

• Understanding evolutionary history relies on tracking population size changes.
• The site frequency spectrum (SFS) summarizes genetic variation within a species.
• Estimating past population dynamics from genetic data is crucial.

Purpose of the Study:

• To introduce CubSFS, a novel method for inferring population size fluctuations in panmictic populations.
• To provide a robust framework for analyzing the site frequency spectrum (SFS) for demographic inference.

Main Methods:

• Derived the expected site frequency spectrum (SFS) based on population size using eigenvalue decomposition.
• Developed an inverse problem solution using cubic splines to model population size changes.
• Optimized the cubic spline fit by minimizing goodness-of-fit and smoothness penalty terms, with weights determined via cross-validation.

Main Results:

• Validated the CubSFS method on simulated demographic histories.
• Applied CubSFS to analyze both unfolded and folded SFS data.
• Successfully inferred demographic histories for 26 diverse human populations from the 1000 Genomes Project.

Conclusions:

• CubSFS offers a powerful new tool for estimating demographic histories from genetic variation data.
• The method accurately reconstructs past population size changes.
• This approach enhances our understanding of human population genetics and evolutionary trajectories.