Robust Design for Coalescent Model Inference.

Area of Science:

• Population genetics
• Evolutionary biology
• Computational biology

Background:

• Coalescent processes model how population size changes affect genetic genealogies.
• Estimating population size changes from genetic data is crucial in biology.
• Existing research on experimental design for coalescent inference is limited and simulation-based.

Purpose of the Study:

• To develop provable and general design theorems for coalescent inference.
• To address key experimental design problems in temporal, spatio-temporal, and time discretization sampling.
• To provide a robust framework for optimizing the estimation of population size changes.

Main Methods:

• Examined temporal sampling under the skyline demographic coalescent model.
• Analyzed spatio-temporal sampling under the structured coalescent model.
• Investigated time discretization for sequentially Markovian coalescent models.
• Proved a robust design theorem based on logarithmic parameter transformation and uniform distribution of coalescent events.

Main Results:

• Demonstrated that working with the logarithm of parameters and uniformly distributing coalescent events minimizes estimation uncertainty.
• Showed this approach is robust, making estimates insensitive to unknown true parameter values.
• Provided rigorous justification for existing experimental design choices in coalescent inference.

Conclusions:

• The derived robust design theorem offers a paradigm for experimental design in coalescent inference.
• Results provide practical guidelines for empirical and simulation-based investigations in population genetics.
• This work bridges the gap between coalescent methodology and experimental design.