A principled approach to deriving approximate conditional sampling distributions in population genetics models with...

Area of Science:

• Population Genetics
• Computational Biology
• Statistical Genomics

Background:

• The multilocus conditional sampling distribution (CSD) is crucial for various population genomics tasks, including phasing, imputation, and ancestry inference.
• Existing CSD approximations, often based on hidden Markov models, lack direct derivation from coalescent theory principles.
• This necessitates the development of more accurate and theoretically grounded CSDs.

Purpose of the Study:

• To develop a principled method for deriving improved multilocus conditional sampling distributions (CSDs).
• To create CSDs that are directly grounded in the underlying population genetics model, specifically the coalescent process with recombination.
• To enhance the accuracy and applicability of CSDs in population genomics.

Main Methods:

• Utilized diffusion process approximation to derive new CSDs from fundamental population genetics models.
• Incorporated intuitive genealogical interpretations of the derived mathematical expressions.
• Developed approximations to ensure scalability with an increasing number of loci.

Main Results:

• The new CSDs are derived directly from the coalescent with recombination model.
• The approach allows for intuitive genealogical interpretations and scalable computations.
• Empirical results demonstrate substantial accuracy improvements over existing CSD approximations.

Conclusions:

• The developed method provides a principled approach to derive accurate multilocus conditional sampling distributions.
• These improved CSDs have significant implications for advancing computational biology and population genomics.
• The method offers a more robust and accurate tool for analyzing DNA sequence data.