Accounting for ambiguity in ancestral sequence reconstruction.

Area of Science:

• Molecular Evolution
• Phylogenetics
• Bioinformatics

Background:

• Ancestral genetic sequence reconstruction is vital for understanding molecular evolution.
• Current methods infer ancestral states but lack information on inference uncertainty.
• Standard criteria maximize or minimize statistical measures, overlooking uncertainty.

Purpose of the Study:

• Introduce a novel criterion, minimum posterior expected error (MPEE), for ancestral sequence reconstruction.
• Address the limitation of existing methods in conveying inference uncertainty.
• Improve the accuracy and precision of reconstructing ancestral genetic sequences.

Main Methods:

• Developed the minimum posterior expected error (MPEE) criterion.
• Evaluated a Brier score-based criterion, independent of tuning parameters.
• Assessed other criteria with arbitrary tuning parameters.
• Conducted large-scale simulations to compare methods.

Main Results:

• MPEE selects a single state when data signal is strong, and multiple states otherwise.
• MPEE and Brier-based criteria show substantial increases in ancestral sequence inference accuracy.
• These new criteria offer realistic compromises on solution precision.
• Outperformed standard approaches in large-scale simulations.

Conclusions:

• The MPEE criterion offers a significant advancement in ancestral sequence reconstruction.
• New criteria improve accuracy by incorporating uncertainty, crucial for molecular evolution studies.
• PhyML software implements MAP and MPEE for nucleotide and amino-acid sequences.