I-Mutant2.0

I-Mutant2.0 predicts the impact of single point mutations on protein stability using support vector machines and accepts either protein sequence or structural input.

Key Features:

• Dual Input Capability: Predicts stability changes from either raw protein sequence or protein structure data.
• SVM-Based Predictions: Uses support vector machines to perform classification of stability change sign (increase/decrease) and regression to estimate ΔΔG values.
• Training Data: Models were trained and validated on experimental thermodynamic data from the ProTherm database.
• Predictive Performance: Cross-validation performance: classification accuracies of 80% with structural input and 77% with sequence input; regression correlations of 0.71 (standard error 1.30 kcal/mol) with structure and 0.62 (standard error 1.45 kcal/mol) with sequence.

Scientific Applications:

• Protein engineering: Prioritizes mutations predicted to stabilize or destabilize proteins for design and optimization workflows.
• Structure-free stability assessment: Enables stability predictions for proteins lacking atomic-resolution structures using sequence-based input.
• Mutational effect interpretation: Supports interpretation of how single point mutations affect protein thermodynamic stability under experimental conditions represented in ProTherm.
• Biopharmaceutical and industrial protein development: Assists in selecting mutations to improve protein stability for therapeutic or industrial applications.

Methodology:

Support vector machines trained on ProTherm experimental data perform both classification of stability change sign and regression of ΔΔG, with performance assessed by cross-validation.