Large-scale prediction of protein geometry and stability changes for arbitrary single point mutations

Area of Science:

• Biochemistry
• Structural Biology
• Computational Biology

Background:

• Predicting the impact of point mutations on protein stability is crucial for protein engineering.
• Existing methods often struggle with diverse mutation types and lack accurate geometry prediction.

Purpose of the Study:

• To develop and validate a novel computational method for predicting protein point mutant geometry and relative stability.
• To assess the accuracy of the prediction method using a large benchmark dataset of experimental stability values.

Main Methods:

• Developed an empirical energy function incorporating folded/denatured protein energies and predicted mutant side-chain conformations.
• Trained the energy function on a large dataset of experimental stability values for single point mutations.
• Validated the method on a separate test set, evaluating prediction accuracy using standard deviation and correlation coefficients.

Main Results:

• The method achieved a standard deviation of 1.08 kcal/mol and a correlation coefficient of 0.82 on the training set.
• On the test set, the prediction yielded a standard deviation of 1.10 kcal/mol and a covariance of 0.66 for 97% of the data.
• Conformation-independent residue contributions were found to account for approximately half of the stability variation.

Conclusions:

• The developed method provides accurate predictions of protein point mutant geometry and stability.
• The findings offer insights into improving protein stability, even without detailed structural information.
• This tool can significantly aid in rational protein design and engineering efforts.