Computing coalescence rates for complex demographies and sampling configurations

Area of Science:

• Population genetics
• Computational biology
• Bioinformatics

Background:

• Population history inference relies on coalescence time distributions.
• Pairwise coalescent rates are popular but limited for recent history.
• Recent coalescences are rare in small sample sizes.

Purpose of the Study:

• Introduce demestats, a software library for computing first-coalescence and cross-coalescence rates.
• Develop statistics with improved power to resolve recent demographic events.
• Enable differentiable demographic models for parameter inference.

Main Methods:

• Compute hazard of first coalescent event for structured demographic models (demes format).
• Combine exact calculations with mean-field approximations for scalability.
• Implement differentiable statistics for parameter optimization.

Main Results:

• Simulations show demestats statistics outperform pairwise summaries for recent population size change and migration.
• Analysis of 1000 Genomes Project data provides new insights into human population expansion rates.
• Successfully applied to tree sequences for demographic inference.

Conclusions:

• Demestats offers a powerful new approach for inferring recent population history from genetic data.
• The library enhances the resolution of demographic inference, particularly for recent events.
• Provides valuable insights into human population dynamics and expansion.