Constructing phylogenetic networks via cherry picking and machine learning.

Area of Science:

• Evolutionary biology
• Computational phylogenetics

Background:

• Phylogenetic tree reconciliation into a single network is a core challenge in evolutionary studies.
• Current methods are computationally intensive, limiting scalability to small datasets or specific network types.

Purpose of the Study:

• To introduce efficient heuristics for constructing phylogenetic networks from multiple input trees.
• To address the computational limitations of existing phylogenetic network inference methods.

Main Methods:

• Application of the 'cherry picking' theoretical framework to develop efficient heuristics.
• Integration of machine learning models to capture tree structure and guide heuristic algorithms.
• Development of fast, randomized heuristics for practical phylogenetic network construction.

Main Results:

• Guaranteed construction of a phylogenetic network encompassing all input trees for binary trees.
• Heuristics demonstrate efficiency on practical-size datasets, outperforming computationally expensive exact methods.
• Machine-learned heuristics provide a novel and promising approach to phylogenetic analysis.

Conclusions:

• The proposed heuristics are scalable to real-world datasets, offering high-quality phylogenetic networks.
• Experimental results validate the effectiveness and accuracy of the developed methods.
• Highlights the potential of machine learning applications in advancing phylogenetics.