Artificial Intelligence Creates Tiny Enzymes for Industrial-Scale Bioprocessing

In a groundbreaking development, a team led by Hiroyuki Hamada has employed ProtGPT2 in an in silico approach to create smaller enzymes for bioprocessing. The focus of their study was malate dehydrogenase (MDH), a crucial enzyme in various biological processes.

The scientists began by collecting amino acid data on MDH and utilising ProtGPT2 to generate sequences that were smaller than the natural version of this enzyme. ProtGPT2, a language model trained on the protein space, is capable of generating de novo protein sequences following the principles of natural ones. It identified functional motifs of MDH and incorporated them into the generated sequences.

Traditional design methods for small proteins often face challenges in stabilising structure and reproducing function. However, ProtGPT2's innovative approach seems to have overcome these hurdles.

The team then analysed the generated sequences, and to their delight, discovered that 9 out of 10 randomly selected sequences from the ProtGPT2-generated enzymes were novel variants. Moreover, these sequences were highly similar to natural MDH sequences, indicating the model's accuracy and efficiency.

InterPro, a database of protein families, domains, and functional sites, revealed active sites in two of the novel sequences generated by ProtGPT2. Furthermore, AlphaFold2, a powerful computational method for predicting protein structures, showed that the 3D structures of these variants resembled structures in natural MDH.

If similar results can be produced with other enzymes, bioprocessors might move towards an in silico-based method of making smaller versions. This could revolutionise the industry, making processes more efficient and cost-effective.

Scientists at Pfizer touted the benefits of using enzymes in bioprocessing more than 20 years ago, and now, with the advent of ProtGPT2, these benefits could be realised on a larger scale. The construction of small proteins is crucial for bioprocessing and drug development, and ProtGPT2 appears to be a significant step forward in this field.

It's worth noting that ProtGPT2 was unveiled by Noelia Ferruz and her colleagues in 2022. While the scientist who revealed ProtGPT2 in 2022, a language model trained on protein structures to generate new protein sequences mimicking natural proteins, is not explicitly named in the provided search results, their work has undoubtedly set a new standard in the field of protein engineering.

This research offers a promising future for bioprocessing, with the potential to design amino acid sequence candidates for small MDHs and, in turn, other enzymes. The implications of this breakthrough extend beyond bioprocessing, potentially impacting various industries that rely on enzymes for their processes.