Accurate Inference of Tree Topologies from Multiple Sequence Alignments Using Deep Learning.

Area of Science:

• Evolutionary Biology
• Computational Biology
• Bioinformatics

Background:

• Phylogenetic tree reconstruction is crucial for understanding species evolution.
• Existing methods have limitations, including accuracy issues in specific parameter spaces and statistical inconsistencies.
• Deep learning offers a promising new avenue for biological research.

Purpose of the Study:

• To develop and evaluate a deep learning model for inferring phylogenetic relationships.
• To assess the accuracy and robustness of the deep learning approach compared to traditional methods.

Main Methods:

• A deep convolutional neural network (CNN) was designed to infer quartet topologies from multiple sequence alignments.
• The CNN was trained using both gapped and ungapped sequence data.
• Performance was evaluated on simulated data and compared against established phylogenetic inference techniques.

Main Results:

• The CNN demonstrated high accuracy in phylogenetic inference on simulated data.
• The deep learning approach proved robust in challenging parameter spaces, such as the Felsenstein and Farris zones.
• CNN-generated confidence scores provided more reliable assessments of topological support than bootstrap or posterior probabilities.

Conclusions:

• Deep learning, specifically CNNs, shows significant potential for improving the accuracy of phylogenetic inferences.
• This approach offers a robust alternative to traditional methods, particularly in complex evolutionary scenarios.
• Further research and practical implementation are warranted to fully leverage deep learning in phylogenetics.