Sequence feature-based prediction of protein stability changes upon amino acid substitutions

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Genetics

Background:

• Amino acid substitutions can destabilize proteins, leading to human diseases.
• Existing machine learning methods for predicting protein stability changes often lack biologically relevant sequence features.

Purpose of the Study:

• To develop a novel machine learning method for predicting protein stability changes based on sequence features.
• To identify the most effective sequence features for accurate prediction of stability changes.

Main Methods:

• Developed a new machine learning approach using Support Vector Machines (SVMs).
• Trained SVM classifiers using experimental data on free energy changes of protein stability upon mutations.
• Evaluated twenty different sequence features for their impact on classifier performance.

Main Results:

• Classifier performance significantly varied depending on the sequence features used.
• The most accurate classifier was achieved using a combination of six specific sequence features.
• This optimal classifier demonstrated an overall accuracy of 84.59%, with 70.29% sensitivity and 90.98% specificity.

Conclusions:

• Sequence features are critical for accurately predicting protein stability changes due to amino acid substitutions.
• High-accuracy predictions can help differentiate between deleterious and tolerant mutations in disease candidate genes.
• A web server, MuStab, has been developed to provide access to this predictive tool for researchers.