Comparing Supervised Learning and Rigorous Approach for Predicting Protein Stability upon Point Mutations in...

Area of Science:

• Computational Biology
• Biophysics
• Drug Discovery

Background:

• Accurate prediction of protein stability changes due to point mutations is vital for drug discovery and personalized medicine.
• Current computational methods, including supervised learning, struggle with accuracy due to data limitations, especially on challenging targets like frataxin and p53.
• Rigorous approaches show promise but are limited by computational cost and handling charge-changing mutations.

Purpose of the Study:

• To compare supervised learning and rigorous approaches for predicting protein stability changes upon point mutations.
• To evaluate performance on difficult targets (frataxin and p53) using blind test sets.
• To develop an improved rigorous method for accurate prediction of both conserving and charge-changing mutations.

Main Methods:

• Comparison of supervised learning-based methods with rigorous alchemical approaches.
• Utilized the pmx double-system/single-box technique for an enhanced alchemical protocol.
• Applied methods to predict folding free energy changes for point mutations in frataxin and p53.

Main Results:

• The rigorous alchemical method significantly outperformed state-of-the-art techniques on the frataxin and p53 blind test sets.
• The enhanced alchemical protocol accurately predicted folding free energy changes for both conserving and charge-changing mutations.
• The improved method demonstrated superior overall performance compared to existing methods.

Conclusions:

• Rigorous alchemical methods, particularly with the enhanced pmx protocol, offer a more accurate approach to predicting protein stability changes from mutations.
• This advancement holds significant potential for improving drug discovery and personalized medicine applications.
• The developed protocol addresses limitations of previous methods in handling diverse mutation types.