Unsupervised genome-wide recognition of local relationship patterns

Area of Science:

• Genomics
• Phylogenetics
• Bioinformatics

Background:

• Discordant phylogenetic relationships can arise from incomplete lineage sorting, horizontal gene transfer, and gene duplication.
• Identifying and characterizing these local phylogenetic variations is crucial for understanding their biological roles.
• Automated, large-scale methods are needed for efficient analysis of these genomic phenomena.

Purpose of the Study:

• To develop and validate Saguaro, a computational tool for characterizing local phylogenetic relationships within aligned sequences.
• To automate the identification of genomic regions with distinct phylogenetic patterns discordant with species phylogeny.

Main Methods:

• Saguaro combines a Hidden Markov Model (HMM) and a Self-Organising Map (SOM) to analyze sequence data.
• It utilizes matrices of pair-wise distance measures (cacti) to model local phylogenetic relationships.
• The SOM iteratively refines hypotheses for new cacti in regions poorly modeled by existing ones.

Main Results:

• Saguaro successfully identified regions under lineage-specific constraint in Dengue virus strains.
• In primate genomes, it detected genomic segments attributed to incomplete lineage sorting.
• Approximately 3% of the primate genome was classified as incompatible with expected species phylogeny, with overlaps in immune system genes.

Conclusions:

• Saguaro effectively detects distinct local phylogenetic relationships without prior hypotheses.
• The tool demonstrated utility across diverse datasets, including viral strains and complex primate genomes.
• Saguaro is applicable to a wide range of populations and is available for download.