fastbaps

fastbaps implements a rapid approximation to a Dirichlet Process Mixture (DPM) model to cluster multilocus genotype data for population-genomic and phylogenetic analyses.

Key Features:

• Dirichlet Process Mixture approximation: Provides a rapid approximation to a DPM model specifically tailored for clustering multilocus genotype data.
• Scalability and speed: Scales to very large datasets and reports speed improvements on the order of 10–100 times faster than prior model-based methods.
• Hierarchical partitioning: Rapidly partitions existing hierarchies to maximize the DPM model marginal likelihood for splitting phylogenetic trees into clades and subclades using population genomic models.
• Empirical demonstration: Applied to datasets including over 110,000 sequences of HIV-1 pol genes to demonstrate handling of large-scale sequence data.
• Performance validation: Tested on simulated data and real-world bacterial and viral datasets, producing solutions comparable or superior to previous methods.

Scientific Applications:

• Population genomics: Clusters multilocus genotype data to identify subpopulations and population structure within species.
• Phylogenetics: Splits phylogenetic trees into clades and subclades to aid analysis of evolutionary relationships.
• Pathogen diversity analysis: Characterizes genetic diversity and substructure in complex viral and bacterial populations.

Methodology:

Performs a rapid approximation to a Dirichlet Process Mixture (DPM) model and applies hierarchical partitioning to maximize the DPM marginal likelihood for splitting phylogenetic trees into clades and subclades using population genomic models.