Structure Based Thermostability Prediction Models for Protein Single Point Mutations with Machine Learning Tools.

Area of Science:

• Protein Engineering
• Computational Biology
• Biophysics

Background:

• Protein thermostability is crucial for engineering applications.
• Predicting the impact of point mutations on protein stability is challenging.
• Databases like ProTherm provide experimental data for model development.

Purpose of the Study:

• To develop accurate in silico models for predicting protein mutant thermostability.
• To guide protein design and mutagenesis strategies.
• To identify key features influencing protein stability.

Main Methods:

• Utilized ProTherm database for protein single point mutation data (ddG, dTm).
• Employed Rosetta for folding free energy change calculations.
• Applied informatics modeling with five supervised machine learning methods and PLS regression.
• Performed feature selection and model evaluation for binary/ternary classification and regression.

Main Results:

• Rosetta-calculated folding free energy change was the most influential feature.
• Machine learning models incorporating structural and physical properties achieved high accuracy.
• Evaluated models for predicting changes in unfolding free energy (ddG) and melting temperature (dTm).

Conclusions:

• In silico prediction of protein thermostability is feasible and valuable for protein engineering.
• A combination of computational energy calculations and machine learning effectively predicts stability changes.
• The developed models can guide focused mutagenesis for desired protein properties.