Predicting the ancestral character changes in a tree is typically easier than predicting the root state

Area of Science:

• Computational Biology
• Phylogenetics
• Bioinformatics

Background:

• Predicting ancestral sequences from homologous sequences is crucial in evolutionary studies.
• High substitution rates can render ancestral state reconstruction at the tree root infeasible.
• Existing methods often focus on root reconstruction, with less attention to internal nodes.

Purpose of the Study:

• To investigate the feasibility of reconstructing ancestral states at internal nodes of a phylogenetic tree.
• To compare the accuracy of ancestral reconstruction at internal nodes versus the tree root.
• To evaluate the robustness of different reconstruction methods.

Main Methods:

• Theoretical analysis of ancestral state reconstruction.
• Information theory applied to phylogenetic trees (e.g., Yule tree).
• Computer simulations using various phylogenetic scenarios and methods (counting, parsimony, likelihood).

Main Results:

• Reconstructing ancestral states at internal nodes can be easier than at the root, depending on tree structure and branch lengths.
• Reconstruction remains feasible for internal nodes even at high substitution rates, particularly for the Yule tree.
• Parsimony and likelihood methods demonstrate significantly higher predictive accuracy for internal nodes compared to the root.
• These methods exhibit robustness against sampling bias and model misspecification.

Conclusions:

• Ancestral state reconstruction is more reliable for internal nodes than the root in phylogenetic analysis.
• Internal node reconstruction offers a promising avenue for evolutionary inference, even under challenging evolutionary conditions.
• Parsimony and likelihood approaches are robust and accurate for internal node reconstruction, enhancing their utility in bioinformatics.