Disentangling coevolutionary constraints for modeling protein conformational heterogeneity.

Area of Science:

• Structural Biology
• Computational Biology
• Bioinformatics

Background:

• Characterizing multi-state protein conformations is vital for understanding protein function and developing targeted therapies.
• Coevolutionary constraints from homologous sequences offer insights into protein structure and function.
• MSA Transformer effectively captures coevolutionary signals using attention mechanisms.

Purpose of the Study:

• To develop a method, EvoSplit, for disentangling coevolutionary signals related to distinct protein conformations.
• To leverage these signals for improved protein structure prediction.
• To identify proteins with potential conformational diversity, particularly those relevant to cancer.

Main Methods:

• Utilizing multi-conformational coevolutionary signals captured by MSA Transformer.
• Applying EvoSplit to disentangle conformation-specific coevolutionary information.
• Evaluating EvoSplit's performance against existing methods like AF-Cluster.
• Predicting conformations for proteins beyond the training set of structure prediction tools.
• Analyzing protein-protein interactions and performing molecular dynamics simulations.

Main Results:

• EvoSplit outperforms AF-Cluster on 85 fold-switching proteins.
• The method successfully models conformations for proteins not included in AlphaFold2's training data.
• 54 candidate human proteins with potential conformational diversity relevant to cancer were identified.
• EvoSplit consistently predicted two conformations for five GTPases, with one being a novel finding.
• Novel HRAS function-associated conformations were revealed through protein-protein interaction analysis.

Conclusions:

• EvoSplit effectively disentangles coevolutionary signals to predict distinct protein conformations.
• This approach enhances protein structure prediction accuracy, especially for dynamic proteins.
• The identification of novel conformations in cancer-related proteins opens avenues for new therapeutic strategies.
• Further validation through evolutionary analysis and molecular dynamics simulations supports the findings.