Scalable Species Tree Inference with External Constraints.

Area of Science:

• Phylogenetics
• Computational Biology
• Bioinformatics

Background:

• Species tree inference is crucial for biological discovery but faces challenges from gene tree discordance and large datasets.
• Existing methods like ASTRAL-J can address gene tree discordance but may not fully leverage available species tree information for speed.
• Computational efficiency is a significant bottleneck in phylogenomic analyses with numerous species and genes.

Purpose of the Study:

• To develop novel methods for species tree inference that utilize partial species tree knowledge.
• To improve computational efficiency and accuracy in species tree estimation, particularly for large datasets.
• To provide statistically consistent methods under the multispecies coalescent model with constraints.

Main Methods:

• Introduction of two new species tree estimation methods: NJst-J and FASTRAL-J.
• These methods incorporate a nonbinary unrooted constraint tree representing partial species tree knowledge.
• Statistical consistency was proven under the multispecies coalescent model with constraints.

Main Results:

• NJst-J and FASTRAL-J demonstrate significantly faster runtimes compared to ASTRAL-J.
• FASTRAL-J maintains accuracy comparable to ASTRAL-J, while NJst-J offers particularly high speed.
• Analysis of the Avian Phylogenomics Project dataset shows substantial reductions in computation time for FASTRAL-J and NJst-J over ASTRAL.

Conclusions:

• NJst-J and FASTRAL-J offer efficient and accurate alternatives for species tree inference, especially when partial tree information is available.
• These methods address key computational challenges in phylogenomics, enabling faster and more scalable analyses.
• The developed methods enhance the utility of phylogenomic data for biological discovery.