IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies.

Area of Science:

• Computational Biology
• Phylogenetics
• Bioinformatics

Background:

• Phylogenomic data sets necessitate rapid tree inference methods for constructing maximum-likelihood (ML) phylogenies.
• Existing fast ML programs may not guarantee finding the optimal tree due to heuristic search strategies.

Purpose of the Study:

• To develop and evaluate novel, time-efficient approaches for ML phylogenetic tree inference.
• To explore alternative search strategies that improve the exploration of tree-space.

Main Methods:

• Implementation of a combined hill-climbing and stochastic perturbation method for tree search.
• Comparative analysis of IQ-TREE against established software (RAxML, PhyML) using large phylogenomic data sets.
• Evaluation of likelihood scores and computational time across different data types (DNA, protein).

Main Results:

• IQ-TREE achieved higher likelihoods than RAxML and PhyML in 62.2%–87.1% of alignments when given equivalent CPU time.
• Using its own stopping rule, IQ-TREE demonstrated improved likelihoods (73.3%–97.1%) compared to faster execution times of RAxML and PhyML on certain alignments.
• IQ-TREE efficiently explores tree-space, offering a valuable alternative for large-scale phylogenetic analyses.

Conclusions:

• The combination of hill-climbing and stochastic perturbation offers a time-efficient strategy for ML phylogenetic inference.
• IQ-TREE provides a robust and efficient tool for large phylogenomic data sets, enhancing the accuracy of phylogenetic tree reconstruction.