Nonbifurcating Phylogenetic Tree Inference via the Adaptive LASSO

Area of Science:

• Computational Biology
• Evolutionary Biology
• Genomics

Background:

• Deep DNA sequencing is revolutionizing evolutionary studies, particularly for rapidly evolving systems like virus-host interactions.
• Phylogenetic trees from dense sampling can exhibit features like sampled ancestors and polytomies, indicating simultaneous or immediate divergence.
• These features are linked to zero-length branches, which are currently undetectable by standard maximum-likelihood methods.

Purpose of the Study:

• To develop a method for identifying zero-length branches in phylogenetic trees.
• To improve the accuracy and detail of phylogenetic inference, especially for rapidly evolving pathogens.
• To enhance the understanding of within-host viral evolution and immune system dynamics.

Main Methods:

• Introduction of adaptive-LASSO-type regularization estimators for phylogenetic tree branch lengths.
• Derivation and analysis of the statistical properties of these novel estimators.
• Demonstration of the practical utility of regularization techniques in phylogenetic analysis.

Main Results:

• Successfully identified zero-length branches in phylogenetic trees, which were previously obscured.
• Adaptive-LASSO regularization proved effective in detecting these critical evolutionary events.
• The proposed method offers a significant advancement over existing maximum-likelihood approaches for phylogenetics.

Conclusions:

• Regularization-based estimation is a powerful and practical approach for uncovering hidden structures in phylogenetic trees.
• This method enhances the resolution of evolutionary histories, particularly in complex and rapidly changing biological systems.
• The findings contribute to a more accurate reconstruction of viral evolution and host immune responses.