mStruct

mStruct models population structure in genetic datasets by integrating an admixture (mixed membership) model with an explicit mutation process to jointly infer ancestry and mutation patterns.

Key Features:

• Admixture model with mutation incorporation: Employs a mixed membership (admixture) model that integrates a mutation process within observed genetic markers to account for both ancestry mixture and allele changes.
• Hierarchical Bayesian framework: Uses a hierarchical Bayesian framework to jointly infer population structure and estimate mutation parameters.
• Variational algorithm for inference: Implements a variational algorithm to perform scalable approximate inference for the joint structure-and-mutation model.

Scientific Applications:

• Evolutionary biology and population genetics: Applied to study population structure and admixture in evolutionary and population-genetic analyses.
• Divergence and migration inference: Aids estimation of divergence times and migration histories of admixed populations by jointly modeling ancestry and mutations.
• Validation and dataset applications: Validated on synthetic datasets and applied to Human Genome Diversity Project-Centre d'Etude du Polymorphisme Humain (HGDP-CEPH) microsatellite and single-nucleotide polymorphism (SNP) datasets, and compared against Structure for assessing structural maps and mutation patterns.

Methodology:

mStruct combines a mixed membership (admixture) model with an explicit mutation process within a hierarchical Bayesian framework and performs approximate inference using a variational algorithm.