pong

pong processes and visualizes post-processing results from mixed-membership clustering of multilocus genotype data to characterize population genetic structure.

Key Features:

• Network-graphical approach: Implements a network-graphical method for analyzing and visualizing membership in latent clusters from clustering inference outputs.
• Assignment Problem solver: Uses algorithms to solve the Assignment Problem for aligning cluster labels across replicate runs and varying numbers of latent clusters (K).
• Support for mixed-membership models: Handles outputs from mixed-membership models analogous to latent Dirichlet allocation and accommodates variable output matrices and K selection.
• Efficient algorithms and scalability: Employs computationally efficient algorithms that reduce runtime and scale to large datasets, demonstrated on 225,705 unlinked genome-wide single-nucleotide variants from 2,426 unrelated individuals in the 1000 Genomes Project.

Scientific Applications:

• Population genetics: Post-processes clustering inference outputs to quantify, visualize, and annotate admixture and cluster membership in population-structure studies.
• Ecology: Applies to ecological datasets analyzed with mixed-membership models to resolve population structure and mixture proportions.
• Text data mining: Supports analysis of mixed-membership model outputs in text corpora where latent Dirichlet allocation–style models are used.

Methodology:

Computational methods explicitly include a network-graphical approach and algorithmic solutions to the Assignment Problem to align cluster labels and summarize mixed-membership inference outputs.