AI designs functional enzyme components that could help expand drug discovery

Researchers at the Helmholtz Institute for Pharmaceutical Research Saarland have demonstrated that artificial intelligence (AI) can aid in the development of novel protein building blocks for biological production systems. These blocks, known as thiolation domains (T-domains), could potentially expand the range of active ingredients available for drug discovery.

The study, published in Nature Communications, focused on non-ribosomal peptide synthetases (NRPS), complex enzyme structures that produce a variety of biologically active compounds. Traditional methods to improve these enzymes have been hindered by the complexity of their structures and the likelihood that modifications result in nonfunctional proteins.

The AI-driven approach developed by Dr. Kenan Bozhüyük's team aimed to overcome this challenge by understanding the underlying biological processes and predicting new T-domain designs. By combining pretrained AI models with experimental testing, the researchers created 76 new T-domains. Subsequently, they tested these in various NRPS variants, observing that several designs were functional and led to increased product yields, sometimes up to three times higher than natural variants.

Notably, some AI-designed T-domains exhibited enhanced solubility and thermal stability compared to their natural counterparts. This success was attributed to the AI's ability to tailor the new building blocks to their specific environment within the enzyme systems, recognizing the importance of local interactions for functionality.

The researchers emphasize that while AI does not replace experimental work, it can efficiently select promising candidates from numerous potential variants. This advancement could significantly accelerate the development of new drugs based on natural products, potentially leading to quicker clinical applications and more efficient drug development processes.