INPS-MD

INPS-MD predicts the impact of single-point non-synonymous mutations on protein stability by combining sequence-based SVM regression and 3D-structure-derived features to estimate thermodynamic free energy changes (ΔΔG).

Key Features:

• Sequence-Based Prediction (INPS): INPS uses Support Vector Machine (SVM) regression to predict thermodynamic free energy change (ΔΔG) upon single-point variations in protein sequences, performs comparably to structure-based methods in cross-validation on non-redundant datasets, and shows high efficacy on datasets such as the tumor suppressor protein p53.
• Structure-Based Prediction (INPS3D): INPS3D incorporates features derived from protein 3D structures and achieves Pearson's correlation of 0.58 with experimental ΔΔG in cross-validation and 0.72 on a blind test set.
• Complementary Predictions: Combining INPS and INPS3D provides complementary insights that improve overall prediction accuracy and enables more comprehensive annotation of mutation effects, notably for proteins like p53.

Scientific Applications:

• Protein Engineering: Predicting stability changes due to mutations to support the design of proteins with altered stability or function.
• Genomic Research: Elucidating molecular mechanisms of human disease by analyzing effects of Mendelian and somatic non-synonymous mutations on protein stability.
• Drug Development: Informing development of therapeutic strategies by assessing how specific genetic variations affect target protein stability.

Methodology:

Applies Support Vector Machine (SVM) regression to sequence-derived features and incorporates structural features from protein 3D structures to predict thermodynamic free energy changes (ΔΔG) upon single-point mutations.