Associating protein sequence positions with the modulation of quantitative phenotypes

Area of Science:

• Computational Biology
• Protein Science
• Bioinformatics

Background:

• Predicting protein function from sequence is challenging.
• Identifying specific amino acid contributions to protein folding and function remains difficult.
• Current methods for predicting protein phenotypes are often complex and data-intensive.

Purpose of the Study:

• To develop a simple algorithm for identifying protein sequence positions that modulate quantitative phenotypes.
• To correlate per-position sequence differences with observed phenotype differences.
• To predict protein functional features using sequence-phenotype relationships.

Main Methods:

• Performed multiple sequence alignments on protein sequences.
• Calculated per-position pairwise differences for sequences and phenotypes.
• Computed correlations between sequence and phenotype differences; applied a linear model for prediction.

Main Results:

• Identified 3 to 10 key positions associated with specific phenotypes across four diverse biological systems.
• Demonstrated that analyzing correlated positions jointly improves prediction accuracy.
• Achieved phenotype predictions comparable to state-of-the-art, more complex methods.

Conclusions:

• The developed algorithm efficiently identifies sequence determinants of biological activity.
• It enables accurate prediction of functional features for uncharacterized proteins.
• This low-information-cost approach is broadly applicable across various protein families.