Inference of complex demographic history using composite likelihood based on whole-genome genealogies

Area of Science:

• Population genetics
• Computational biology
• Bioinformatics

Background:

• Accurate parametric inference for complex demographic models remains a challenge in population genetics.
• Ancestral recombination graphs (ARGs) offer richer data than summary statistics for demographic inference.

Purpose of the Study:

• To introduce mrpast, a novel computational tool for inferring complex demographic histories from ARGs.
• To evaluate the accuracy and versatility of mrpast across diverse population genetic models.

Main Methods:

• mrpast utilizes a composite likelihood approach based on pairwise sample coalescence times from ARGs.
• Employs a continuous-time Markov process for coalescence probabilities.
• Evaluated on simulated and inferred ARGs for various demographic scenarios.

Main Results:

• mrpast accurately infers demographic parameters including split times, population sizes, and gene flow.
• Demonstrated high accuracy on complex models like asymmetric migration, population size changes, and admixture.
• Successfully reconstructed American admixture history using a "ghost" population approach.

Conclusions:

• mrpast provides a robust and versatile pipeline for complex demographic inference from ARGs.
• The tool facilitates simulation, inference, and manipulation of ARGs for model testing.
• mrpast enhances the accuracy and ease of population genetic demographic inference.