SAAFEC-SEQ: A Sequence-Based Method for Predicting the Effect of Single Point Mutations on Protein Thermodynamic...

Area of Science:

• Computational biology
• Biophysics
• Machine learning in bioinformatics

Background:

• Predicting protein thermodynamic stability changes from mutations is crucial for protein engineering and understanding disease mechanisms.
• Existing sequence-based methods often lack accuracy or require structural data, limiting their application.
• Genome-scale analyses are hindered by the scarcity of protein structural information.

Purpose of the Study:

• To develop and validate SAAFEC-SEQ, a novel machine learning method for predicting the change in protein folding free energy due to amino acid substitutions.
• To create a sequence-only method applicable to large-scale genomic studies.
• To establish a new state-of-the-art benchmark for sequence-based stability prediction.

Main Methods:

• Utilized a gradient boosting decision tree machine learning approach.
• Incorporated physicochemical properties, sequence features, and evolutionary information.
• Developed SAAFEC-SEQ, a method requiring only the protein sequence, not its 3D structure.

Main Results:

• SAAFEC-SEQ consistently outperformed existing state-of-the-art sequence-based methods.
• Performance was validated using Pearson correlation coefficient and root-mean-squared-error metrics on multiple independent datasets.
• The method demonstrated high accuracy in predicting the effects of mutations on protein stability.

Conclusions:

• SAAFEC-SEQ represents a significant advancement in predicting mutation-induced changes in protein thermodynamic stability.
• The sequence-based nature of SAAFEC-SEQ enables broad applicability, including genome-scale investigations.
• The tool is accessible via a web server and downloadable code, facilitating its integration into diverse research workflows.