Site-Specific Amino Acid Distributions Follow a Universal Shape

Area of Science:

• Computational Biology
• Evolutionary Biology
• Biophysics

Background:

• Protein evolution models are crucial for evolutionary inference.
• Existing models are either too simplistic (e.g., dN/dS) or too complex (e.g., mutation-selection models).
• A need exists for models balancing biophysical realism and parameter parsimony.

Purpose of the Study:

• To investigate if a middle ground exists between coarse-grained and complex protein evolution models.
• To determine if site-specific amino acid frequencies can be realistically described with a limited number of parameters.
• To explore a novel approach for modeling amino acid variation.

Main Methods:

• Developed a novel distribution model with a single free parameter.
• Applied the model to analyze amino acid variation in multiple sequence alignments.
• Compared model predictions with empirical protein sequences and simulated data from all-atom force fields.
• Focused on predicting amino acid frequencies by rank order.

Main Results:

• A single-parameter distribution model accurately captures amino acid frequency variation at most sites.
• This accuracy is maintained whether analyzing empirical or simulated protein sequences.
• The model's success relies on predicting frequency ranks, not specific amino acid identities.
• A near-universal shape was observed in amino acid frequency distributions.

Conclusions:

• A simplified, rank-based approach can yield realistic site-specific amino acid frequency predictions.
• This finding suggests a new direction for developing more realistic yet parsimonious evolutionary models.
• The universal shape of frequency distributions offers a fundamental insight into protein evolution.