Redesigning myoglobin via functional site scaffolding for enhanced catalytic functions

Protein redesign is frequently limited by the scarcity of stable and robust scaffolds. While computational methods can expand protein sequence space to generate novel scaffolds, reproducing the fine structural features essential for the function of metalloproteins like myoglobin remains challenging. In this work, we employed deep learning–based functional site scaffolding strategy, in which backbone architectures were generated using a diffusion-based structural model, sequences were optimized through inverse-folding design (ProteinMPNN), and structural consistency was subsequently evaluated using structure prediction algorithms (AlphaFold and OmegaFold), initiating the design process from the heme-binding pocket and secondary coordination sphere residues of myoglobin. Through the generation and computational screening of over 100,000 sequences, we obtained a miniaturized version of myoglobin, termed bitMb, that preserves the key heme-binding features as well as the native O₂ binding behaviour. To assess its catalytic potential, we introduced known beneficial mutations from sperm whale myoglobin into bitMb. The resulting variants demonstrated increased peroxidase and carbene transferase activities compared with the parent bitMb scaffold, confirming the scaffold's functional flexibility. Additionally, bitMb exhibited enhanced thermal stability, with a melting temperature 5.5 °C higher than that of the wild-type sperm whale myoglobin, and remarkable stability in high concentrations of organic solvents, retaining heme-binding capability and enzymatic activity in up to 96.7% methanol. Our results demonstrate that the functional site scaffolding-based redesign strategy can generate robust and versatile protein scaffolds capable of diverse catalytic functions.