Design thermostable proteins (HyperMPNN)

HyperMPNN is a deep learning-based inverse folding model designed specifically to enhance protein thermostability. Built upon ProteinMPNN, it was retrained on predicted protein structures from hyperthermophilic organisms (optimal growth temperatures up to 105°C), learning the thermotolerance strategies refined by nature over billions of years of evolution.

Unlike general-purpose models, HyperMPNN accurately recapitulates the characteristic amino acid composition preferences of thermophilic proteins — charged residues enriched on the surface, hydrophobic residues enriched in the core — and successfully transfers these design principles to proteins of non-thermophilic origin. Experimental validation shows that the I53-50 protein nanoparticle designed by HyperMPNN exhibits a melting temperature increase from 65°C to above 95°C, a stability improvement exceeding 30°C.

This online tool provides convenient protein thermostability design services, suitable for enzyme engineering, vaccine delivery carrier development, biopharmaceutical preservation, and other applications requiring high-temperature tolerance. Simply upload the backbone structure of your target protein to obtain thermostability-optimized candidate sequences and accelerate your R&D work.

Upload a protein structure file in PDB or mmCIF format, and the system will automatically parse chain information. After selecting chains to design and setting parameters, the model will generate multiple optimized amino acid sequences, with scores (score = -log_prob, lower is better) and sequence recovery rates for each.