Predicting protein thermal stability changes upon point mutations using statistical potentials: Introducing HoTMuSiC.

Area of Science:

• Protein Science
• Biotechnology
• Computational Biology

Background:

• Accurate prediction of amino acid substitution effects on protein thermal stability is vital for protein engineering.
• Optimizing enzymes for industrial bioprocesses often requires enhancing their stability under demanding conditions.

Purpose of the Study:

• To develop and validate a computational tool for predicting the change in melting temperature (ΔTm) upon point mutations.
• To provide a user-friendly webserver for predicting protein thermal stability alterations.

Main Methods:

• Utilized standard and temperature-dependent statistical potentials combined with an artificial neural network.
• Developed the model based on a detailed thermodynamic analysis of protein systems.
• Trained and validated the model using over 1,600 experimentally determined mutations with measured ΔTm values.

Main Results:

• Achieved a root mean square deviation (RMSD) of 4.2°C between predicted and experimental ΔTm values.
• Reduced the RMSD to 2.9°C when removing the top 10% of outliers.
• Demonstrated superior performance compared to existing computational methods through rigorous 5-fold cross-validation.

Conclusions:

• The developed computational tool provides accurate predictions of protein melting temperature changes due to mutations.
• The method, integrating statistical potentials and neural networks, offers a significant advancement in predicting protein thermal stability.
• A freely accessible webserver is available for non-commercial use, facilitating protein engineering and bioprocess optimization.