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

In a groundbreaking leap forward for synthetic biology, researchers are harnessing artificial intelligence to design proteins that have never existed in nature. This innovation marks a pivotal moment in the decades-long pursuit of creating novel proteins, which hold transformative potential for medicine, biotechnology, and scientific discovery. Jason Zhang, an assistant professor of bioengineering at the UCLA Samueli School of Engineering, is leading efforts to push the boundaries of protein design using advanced AI techniques. His work builds upon the pioneering research of David Baker, a Nobel laureate in Chemistry whose contributions to computational protein engineering laid the foundation for this new frontier. Zhang previously worked under Baker’s guidance at the University of Washington, where he gained expertise in rational protein design, a method that contrasts with traditional approaches reliant on evolutionary selection. The process begins with the creation of AI-driven models capable of generating vast libraries of potential protein structures. These models are trained on extensive datasets derived from experimental results, enabling them to predict which amino acid sequences will fold into functional proteins. Once generated, these sequences are translated into DNA and synthesized in a laboratory setting, where their functionality is tested against specific targets. This approach allows scientists to evaluate thousands of candidate proteins simultaneously, significantly accelerating the pace of discovery compared to conventional methods. Zhang’s research encompasses a wide range of applications, from developing diagnostic tools to advancing therapeutic strategies. One notable project involves the design of biosensors that can monitor complex biological processes within living systems. These sensors provide detailed insights into cellular functions, potentially paving the way for personalized medicine and more effective treatments for diseases ranging from cancer to neurological disorders. A particularly ambitious goal is the development of a “virtual cell,” a digital representation of biological systems that can simulate cellular behavior and interactions. By integrating large amounts of data collected from engineered biosensors, researchers aim to construct comprehensive models of cellular activity. Such models could enable the prediction of how different drugs interact with specific cell types, offering unprecedented precision in drug development and treatment planning. Collaboration plays a crucial role in translating theoretical advancements into practical applications. Zhang works alongside colleagues at UCLA to develop targeted therapies, including innovative approaches to treat rare forms of liver cancer using CAR-T cell technology, a strategy that has shown remarkable success in treating blood cancers. These collaborative efforts highlight the interdisciplinary nature of modern biomedical research, where computational modeling meets experimental validation. Another critical focus area within Zhang’s lab is the study of disordered proteins, which are implicated in numerous diseases, including neurodegeneration, diabetes, and certain cancers. Traditional small-molecule drugs often struggle to target these proteins due to their lack of defined structure. To overcome this challenge, Zhang’s team employs generative AI to design entirely new protein folds that can specifically bind to disordered proteins. This strategy seeks to transform inherently "undruggable" targets into viable therapeutic candidates by inducing structural changes that make them susceptible to conventional drug mechanisms. As the field continues to evolve, the integration of AI into protein engineering promises to revolutionize our understanding of biological systems and expand the horizons of medical science. With ongoing research and collaboration, the potential for discovering and utilizing proteins never before seen in nature is becoming increasingly tangible.