PPanGGOLiN

PPanGGOLiN partitions and models microbial pangenomes using graph-based representations and statistical classification to identify persistent, shell, and cloud gene families across isolate genomes, metagenome-assembled genomes (MAGs), and single-cell amplified genomes (SAGs).

Key Features:

• Graph-Based Pangenome Modeling: Represents pangenomes as graphs with nodes corresponding to gene families and edges denoting genomic neighborhoods.
• Statistical Partitioning: Uses an Expectation-Maximization algorithm on a multivariate Bernoulli Mixture Model coupled with a Markov Random Field to classify gene families into persistent, shell, and cloud partitions.
• Topology- and Presence/Absence-Aware Classification: Integrates graph topology and gene presence/absence patterns to assign gene families to partitions.
• Support for Low-Quality Genomes: Applies the statistical framework to low-quality data types including metagenome-assembled genomes (MAGs) and single-cell amplified genomes (SAGs).
• Genome Dynamics Insight: Enables analysis of partitioned pangenome graphs to highlight the role of the shell genome and its contribution to species adaptation independent of genome size.

Scientific Applications:

• Comparative Genomics: Enables comparative analyses of gene content and organization across hundreds of microbial species using compact pangenome graphs.
• Functional and Evolutionary Inference: Supports inference of functional repertoires and evolutionary dynamics by distinguishing persistent, shell, and cloud gene families.
• Epidemiological and Diversity Studies: Facilitates epidemiological investigations and studies of microbial diversity using partitioned pangenomes from isolate genomes and MAGs.

Methodology:

Models pangenomes as graphs (nodes = gene families; edges = genomic neighborhoods); applies an Expectation-Maximization algorithm on a multivariate Bernoulli Mixture Model coupled with a Markov Random Field to infer partitions and optimal class numbers while accounting for genome organization and presence/absence patterns; validated on isolate genomes from 439 species and metagenome-assembled genomes from 78 species.