MPTherm-pred

MPTherm-pred predicts changes in thermal stability (ΔTm) induced by mutations in membrane proteins to quantify and interpret mutation effects in membrane-spanning and solvent-exposed regions.

Key Features:

• Dataset: Built on 929 experimentally determined mutations in membrane proteins, including 190 stabilizing and 232 destabilizing mutations within membrane-spanning regions.
• Quantitative ΔTm statistics: For membrane-spanning mutations the average ΔTm is 2.43°C (±3.1°C) for stabilizing and −5.48°C (±5.5°C) for destabilizing mutations, and for solvent-exposed regions the values are 2.56°C (±2.82°C) and −6.8°C (±7.2°C), respectively.
• Determinants of stability: Systematically identifies changes in hydrophobicity, Cα atom contacts, and aliphatic residue frequency as critical determinants for mutations within membrane-spanning regions.
• Machine learning predictors: Employs structure- and sequence-based machine learning algorithms to predict ΔTm, achieving correlation 0.72 with MAE 2.85°C for membrane-spanning mutations and correlation 0.73 with MAE 3.7°C for aqueous-region mutations.
• Validation: Performance validated on a test set of mutations in evolutionarily independent protein sequences to assess robustness and generalizability.

Scientific Applications:

• Membrane protein engineering: Design and selection of mutations to enhance thermal stability of membrane proteins.
• Mechanistic analysis: Investigate molecular determinants of ΔTm changes, including hydrophobicity, Cα contacts, and aliphatic residue frequency.
• Structural biology and biotechnology: Prioritize stable variants for structural and functional studies and biotechnological applications.
• Pharmaceutical development: Support identification of stable membrane protein variants relevant to drug discovery.

Methodology:

Uses structure- and sequence-based machine learning models trained on 929 experimentally determined mutations; analyzes hydrophobicity changes, Cα atom contact counts, and aliphatic residue frequency; evaluates model performance by correlation and MAE on a test set of evolutionarily independent sequences.