INPS: predicting the impact of non-synonymous variations on protein stability from sequence

Area of Science:

• Bioinformatics
• Computational Biology
• Protein Engineering

Background:

• Accurate prediction of protein stability changes is vital for understanding genetic mutations and protein engineering.
• Next Generation Sequencing generates vast amounts of protein sequence data, necessitating tools that do not rely on 3D structures.
• There is a need for sequence-based methods to annotate the effects of mutations.

Purpose of the Study:

• To introduce INPS, a novel sequence-based approach for predicting the impact of non-synonymous mutations on protein stability.
• To evaluate INPS's performance against existing structure-based methods.
• To demonstrate the utility of INPS for analyzing variations in proteins like p53.

Main Methods:

• INPS utilizes Support Vector Machine (SVM) regression.
• The model is trained to predict the change in thermodynamic free energy (ΔΔG) upon single-point variations.
• The method operates directly on protein sequences, without requiring structural information.

Main Results:

• INPS demonstrates performance comparable to state-of-the-art structure-based methods in cross-validation.
• INPS shows strong performance on a dataset of variations in the p53 tumor suppressor protein.
• Combining INPS with the structure-based method mCSM yields improved prediction accuracy for the p53 dataset.

Conclusions:

• INPS is a valuable tool for assessing the effect of non-synonymous polymorphisms on protein stability when structural data is unavailable.
• INPS predictions are complementary to structure-based methods, enhancing overall accuracy when combined.
• The developed method is accessible as a web server.