nhPhyML

nhPhyML implements the non-stationary, non-homogeneous model of DNA sequence evolution (Galtier and Gouy 1998) to compute maximum-likelihood phylogenetic trees under non-reversible substitution processes.

Key Features:

• Non-reversible likelihood computation: Computes tree likelihoods under non-reversible substitution models and can evaluate the likelihood of a tree from any of its nodes.
• Reorganized likelihood algorithm: Reorganizes likelihood computation to enable the application of heuristics for efficient maximum-likelihood estimation on non-reversible models.
• PhyML integration: Adapts the non-stationary, non-homogeneous model to the PhyML framework to explore tree topologies and estimate parameters within that engine.
• Equilibrium frequency approximation: Implements an approximation that limits the set of equilibrium nucleotide frequencies to reduce computational complexity.
• Topology exploration and ancestral composition: Provides robust exploration of tree topology space and supports estimation of ancestral G+C content for complex evolutionary scenarios.

Scientific Applications:

• rRNA site analysis: Analyze slowly evolving sites in rRNA sequences using non-stationary, non-homogeneous models.
• Broad taxonomic sampling: Perform phylogenetic inference across broader taxonomic sampling without assuming reversible substitution processes.
• LUCA hypotheses: Investigate hypotheses about the last universal common ancestor, including the hyperthermophilic hypothesis.
• Ancestral G+C content estimation: Estimate ancestral G+C content in genomes to study compositional evolution.

Methodology:

nhPhyML reorganizes likelihood computation to accommodate non-reversible models, applies heuristics to make maximum-likelihood estimation tractable for large datasets, implements an approximation that restricts equilibrium frequency sets, and adapts the model to PhyML's computational framework.