Inference of Phylogenetic Networks From Sequence Data Using Composite Likelihood.

Area of Science:

• Evolutionary Biology
• Computational Biology
• Genomics

Background:

• Phylogenetic trees are insufficient for depicting evolutionary processes like hybridization.
• Hybridization, the interbreeding of species, requires phylogenetic networks for accurate representation.
• Current methods for phylogenetic network inference are computationally intensive and limit practical application.

Purpose of the Study:

• To introduce PhyNEST (Phylogenetic Network Estimation using SiTe patterns), a novel, scalable method for estimating phylogenetic networks.
• To enable direct inference of binary, level-1 phylogenetic networks from sequence data.
• To overcome the computational limitations of existing phylogenetic network estimation techniques.

Main Methods:

• PhyNEST utilizes composite likelihood inference, processing full genomic data efficiently.
• The method employs hill climbing and simulated annealing algorithms to explore network space.
• Assumptions include coalescent independent sites, Jukes-Cantor substitution model, and constant effective population size.

Main Results:

• PhyNEST demonstrates higher accuracy compared to existing composite likelihood methods (SNaQ and PhyloNet) in simulation studies.
• The method is robust to certain model misspecifications, such as using a simpler substitution model.
• PhyNEST was successfully applied to reconstruct evolutionary relationships in Heliconius butterflies and Papionini primates.

Conclusions:

• PhyNEST offers a computationally tractable and accurate solution for inferring phylogenetic networks directly from sequence data.
• The method facilitates a more comprehensive understanding of evolutionary histories shaped by hybridization and introgression.
• PhyNEST is available as an open-source Julia package, promoting wider adoption in evolutionary research.