GHIST 2024: The 1st Genomic History Inference Strategies Tournament.

Area of Science:

• Population genetics
• Evolutionary biology
• Computational genomics

Background:

• Evaluating population genetic inference methods is complex due to evolutionary history intricacies, model misspecification, and bias.
• Standardized benchmarks are needed to rigorously assess the accuracy of these methods.

Purpose of the Study:

• To establish a community-driven competition, the Genomic History Inference Strategies Tournament (GHIST), for evaluating population genomic inference methods.
• To compare the performance of diverse computational approaches on simulated demographic scenarios of varying complexity.

Main Methods:

• The inaugural GHIST competition (July-November 2024) presented four challenges: bottleneck, split with isolation, secondary contact with complexity, and archaic admixture.
• Participants inferred demographic parameters from error-free VCF files, with performance scored by relative root mean squared error.
• Approximately 60 participants employed a variety of computational strategies.

Main Results:

• Methods utilizing site frequency spectra demonstrated strong performance across challenges.
• Flexible model-building approaches showed advantages in accurately inferring complex demographic histories.
• Performance varied depending on the specific demographic scenario and method employed.

Conclusions:

• GHIST provides a crucial standardized benchmark for assessing population genetic inference tools.
• The competition highlighted the strengths of current methods and identified areas for future development, particularly for complex evolutionary scenarios.
• The ongoing GHIST initiative aims to further advance the reliability of inferring evolutionary history from genomic data.