Ambiguity Coding Allows Accurate Inference of Evolutionary Parameters from Alignments in an Aggregated State-Space

Area of Science:

• Evolutionary Biology
• Computational Biology
• Biophysics

Background:

• Understanding protein evolution requires models that capture genetic variation and functional constraints.
• Practical approaches often balance data availability with useful parameter estimation, such as selection strength or ancestral structure.
• Limited data resolution can hinder the application of advanced evolutionary models.

Purpose of the Study:

• To demonstrate a method for obtaining accurate evolutionary parameter estimates from data with limited resolution.
• To show how to infer ancestral protein states and evolutionary parameters when complete data is unavailable.
• To improve ancestral reconstruction and the estimation of selection strength using adapted substitution models.

Main Methods:

• Encoding observed characters as ambiguous representations within a larger state-space.
• Applying established methods for handling missing data to protein sequence alignments.
• Utilizing adapted codon models (e.g., 61-state) and empirical models (e.g., 55-state) for amino acid data.

Main Results:

• Accurate and unbiased estimation of the selection strength parameter omega ($\omega$) using an adapted 61-state codon model.
• Successful inference of ancestral amino acid side chain configurations using a 55-state model on 20-state amino acid data.
• Significantly improved ancestral reconstruction accuracy by incorporating structural information into even a small fraction of sequences.

Conclusions:

• A novel strategy allows the recovery of crucial evolutionary information from protein sequences with previously inaccessible data.
• This ambiguity-coding approach expands the applicability of sophisticated evolutionary models to lower-resolution sequence data.
• The methods presented enhance our ability to reconstruct protein evolutionary history and understand natural selection.