Q&A: Deploying AI to create proteins never before seen in nature

The article discusses the advancements in artificial intelligence (AI) for designing new proteins never seen in nature. Proteins, as essential components of life, can serve various functions such as structural support, messengers, catalysts, or transport systems depending on their amino acid composition and 3D shape. This versatility makes designed proteins a potential game-changer in medicine, industry, and research.

Jason Zhang, an assistant professor of bioengineering at UCLA Samueli School of Engineering, is a key figure in AI-enabled protein design. Prior to joining UCLA in 2025, he was a postdoctoral researcher at the University of Washington laboratory led by David Baker, who was awarded the 2024 Nobel Prize in Chemistry for breakthroughs in computational protein engineering. Zhang aims to continue the pathbreaking tradition of protein engineering to discover new drugs and gain fundamental biological insights.

In an interview, Zhang shares his excitement about the potential of AI in protein engineering. He recalls how his Ph.D. research involved traditional protein engineering methods, which were laborious processes even when he finished his degree in 2020. However, after seeing studies from David Baker's lab, Zhang discovered the rational design approach, where proteins are engineered rationally rather than relying on evolution.

When Zhang joined Baker's lab, AI was rapidly advancing, and the team developed AI models to improve the success rates of designing functional proteins that could be used as drugs.

The AI models are trained using experimental data, starting with a target protein that the designed proteins should bind to. The AI then generates different protein sequences, which are reverse translated into DNA and expressed in the laboratory. By testing 20,000 designed proteins in a single experiment, researchers can determine whether the proteins successfully bind to the target. Following successful initial tests, the proteins are further validated through experiments in cells and disease models.

Zhang emphasizes that engineering proteins is a means to an end, as he seeks to understand cells better. By building tools like biosensors, the lab aims to measure various biological processes and answer long-standing questions with new perspectives. Zhang's lab also envisions creating a virtual cell by generating large datasets using new tools such as biosensors, potentially leading to the creation of a digital twin.

This digital twin could predict how various drugs affect specific cells, such as immune cells or cancer cells.

One of the lab's current focuses is on disordered proteins, which play a role in neurodegenerative diseases, diabetes, and some cancers. Disordered proteins are challenging to target because they can adopt many different shapes. Zhang's lab uses generative AI to create entirely new protein folds that can bind to disordered proteins, essentially providing a pocket for the disordered protein to fit into and become more ordered. This approach aims to drug the undruggable disordered proteins.

Another significant vision for AI protein design is its potential in precision medicine. As understanding of diseases becomes more granular, with diseases being broken down into various forms, AI protein design could enable personalized therapies. Zhang envisions a future where their methods could create tailored therapies for each patient, potentially revolutionizing the field of medicine.