Evolutionary Constraints Guide AlphaFold2 in Predicting Alternative Conformations and Inform Rational Mutation...

Area of Science:

• Structural Biology
• Computational Biology
• Bioinformatics

Background:

• Understanding protein structural variability is crucial for elucidating biological functions.
• AlphaFold2 excels at static structure prediction but misses dynamic functional states.
• Existing methods for generating conformational ensembles lack interpretability and evolutionary insights.

Purpose of the Study:

• To improve the generation of protein conformational ensembles.
• To identify sequence patterns driving alternative protein fold predictions.
• To integrate evolutionary signals into structural ensemble generation.

Main Methods:

• Developed a refined clustering strategy combining protein language model representations with hierarchical clustering.
• Applied the strategy to generate diverse sequence ensembles for protein families.
• Utilized direct coupling analysis (DCA) on clustered alignments to identify coevolutionary signals.
• Designed and validated stabilizing mutations using molecular dynamics and alchemical free energy calculations.

Main Results:

• Successfully identified high-confidence alternative protein conformations.
• Generated abundant sequence ensembles, enabling robust direct coupling analysis (DCA).
• Uncovered key coevolutionary signals linked to specific protein folds.
• Validated designed mutations for stabilizing distinct conformations.

Conclusions:

• The refined clustering strategy enhances the interpretability of conformational ensembles.
• The method effectively captures diverse protein conformational changes, including fold-switching.
• Integrating evolutionary signals provides a powerful framework for understanding and manipulating protein dynamics.