In its fourth year, the research seed grant program at the Beckman Institute for Advanced Science and Technology will fund one research project beginning in 2026.

The selected research team will create a Center for Orchestrated Agentic Biology. It will develop an AI framework to model molecular pathways involved in brain aging. The seed funding will build the computational and experimental foundation for this effort, exploring how nerve cells function as they age.

“The Beckman Seed Grant Program enables Beckman researchers to work outside their comfort zones and develop outside-the-box projects that push scientific boundaries,” said Beckman Director Steve Maren. “Innovative new teams like this one are thinking about problems in unique ways, in this case, laying the groundwork for a deeper understanding of brain aging.”

In recent years, artificial intelligence has paved the way for innovative scientific research. From predicting the structures of proteins to running molecular-level simulations in real time, AI models have shown potential for significant scientific advancements in a short period.

However, even though these highly specialized models are constantly learning and improving, researchers note that individual models often cannot answer complex scientific questions on their own.

The Beckman research team, led by Haohan Wang, Nien-Pei Tsai and Aleksei Aksimentiev, has proposed a new kind of biological AI framework known as orchestrated agentic biology, which would use existing models together to allow for more comprehensive scientific predictions.

 “Our goal is to move beyond using AI as a collection of individual prediction tools and instead create a system in which specialized models can work together to reason about complex biological questions,” Wang said. “By orchestrating these models, we hope to uncover mechanisms of cellular aging that would be difficult to identify using any single model alone.”

Wang is an assistant professor of information sciences. Tsai is a professor of molecular and integrative physiology and neuroscience. Aksimentiev is a professor of bioengineering and physics as well as the John Bardeen Faculty Scholar for physics.

The team includes Yurui Li and Haochuan Wang, both doctoral students in information sciences. Gabe McKenna, a doctoral student in molecular and cellular biology, is also on the team.

Together, the researchers will use OAB to analyze a cellular circuit known as the E2F-p16INK4a axis. This is an important biological pathway in cell aging that leads to a gradual breakdown of neuron cells as time increases, called senescence.

The research plan is divided into three smaller projects among each of the three principal investigators.

The first will involve creating an AI agent that can break down complex questions about biomechanics. Researchers will have other pre-existing models analyze those smaller questions. AI agents are autonomous software systems that perceive their environments, make independent decisions and execute multi-step actions using specialized tools.

Wang will lead this section, and his team will guide the AI agent so all existing models are able to communicate and work simultaneously, even if other researchers have trained them in the past.

The second, led by Aksimentiev, will primarily focus on creating a structural grounding module. This system takes AI-predicted protein structures and uses molecular-level simulations to determine how these regulatory proteins may bind to the target sequences. Running these simulations will allow researchers to directly test whether the mechanisms proposed by OAB are consistent with how the basic kinetics of the targeted sequences and cells operate.

Tsai’s group will handle the third part of the project. Using the OAB’s predictions, the group will run experiments with both young and aged neurons, measuring protein responses and observing the biomechanics at work in real time. If the OAB predictions are consistent with the experimental results, OAB will be considered valid. If not, the hypothesis graph will be reworked, and the next iteration will be tested.

The team plans to combine expert knowledge of AI systems, molecular physiology and structural biophysics to create a biological and technological template. This template not only provides insight into the circuit being studied but may be reused and reapplied to nearly any cellular pathway.