Regularized sequence-context mutational trees capture variation in mutation rates across the human genome.

Area of Science:

• Population genetics
• Genomics
• Bioinformatics

Background:

• Germline mutations are the source of genetic variation.
• Sequence context influences mutation rates, but existing models face limitations like data sparsity and lack of regularization.
• Accurate mutation rate models are crucial for population genetics.

Purpose of the Study:

• To develop a robust and accurate model for estimating sequence-context dependent polymorphism probabilities.
• To address limitations of previous models, including data sparsity and lack of uncertainty quantification.
• To provide a computational tool for analyzing germline mutation patterns.

Main Methods:

• Developed Baymer, a regularized Bayesian hierarchical tree model.
• Implemented an adaptive Metropolis-within-Gibbs Markov Chain Monte Carlo sampling scheme.
• Applied the model to human population data (1000 Genomes Phase 3) and great ape species.

Main Results:

• Baymer accurately infers polymorphism probabilities and provides well-calibrated posterior distributions.
• The model robustly handles data sparsity and produces parsimonious models.
• Demonstrated shared context-dependent mutation rate architecture across species, enabling transfer learning.

Conclusions:

• Baymer is an accurate and efficient algorithm for estimating polymorphism probabilities, adapting to data sparsity.
• The model enhances understanding of sequence context effects on germline mutations.
• Facilitates comparative genomics and provides a foundation for improved mutation modeling.