The solution surface of the Li-Stephens haplotype copying model

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• The Li-Stephens (LS) haplotype copying model is fundamental to statistical inference in genetics.
• It models chromosomes as imperfect mosaics of population chromosomes.
• The model's parameters, mutation rate (μ) and recombination rate (ρ), lack clear biological interpretation.

Purpose of the Study:

• To re-evaluate the LS model parameters (μ and ρ) as tuning parameters rather than fixed biological rates.
• To develop an efficient algorithm for analyzing the impact of these parameters on LS model outputs.
• To explore the implications for common bioinformatic tasks like genotype imputation and phasing.

Main Methods:

• Developed an algorithm to partition the parameter space ([Formula: see text] and [Formula: see text]) into regions of constant LS output.
• Efficiently enumerated all possible solutions for the LS model.
• Extended the approach to the diploid LS model for genotype phasing.

Main Results:

• The algorithm efficiently identifies regions of constant output in the parameter space.
• Demonstrated the impact of varying [Formula: see text] and [Formula: see text] on LS model performance.
• Found that conventional population-scaled values for [Formula: see text] and [Formula: see text] are near-optimal for imputation.

Conclusions:

• Reinterpreting [Formula: see text] and [Formula: see text] as tuning parameters provides a clearer understanding of the LS model's behavior.
• The developed algorithm efficiently analyzes LS model outputs and parameter effects.
• Standard parameter values may lead to suboptimal performance in phasing, specifically inflating switch error.