Our bodies’ computational power doesn’t just reside in the brain. Inside every living cell, a silent, intricate form of computing takes place as chemical networks process information and make critical decisions.
A new study by University of Chicago postdoctoral fellow Carlos Floyd in collaboration with Prof. Aaron Dinner, Assoc. Prof. Arvind Murugan and Prof. Suriyanarayanan Vaikuntanathan sheds light on this hidden world of “biological hardware,” exploring how cells perform complex feats of decision-making. This process relies on chemical networks that are highly precise, allowing cells to filter through a flood of signals and make the correct choices that determine their function and survival—whether that’s to grow, divide or perform a specific function.
For example, stem cells use signals from the environment to become a specific cell type, like a nerve cell or skin cell.
“This decision needs to be performed without a brain—instead, cells rely on internal chemical reaction networks to compute the correct differentiation pathway,” Floyd explained.
The scientists said the knowledge could form the basis of new technologies, such as “smart” chemical processes.
Overcoming the limit
By training models of chemical reaction networks to perform classification tasks, the team discovered a “thermodynamic” constraint that prevented certain networks from recognizing all but the simplest of patterns.
The laws of thermodynamics are typically used to constrain the energy efficiency of engines, but here, researchers found a surprising new application. They discovered these laws also limit a biological system’s ability to compute—a constraint directly linked to the energy the system can consume.
However, the team found this limitation could be overcome by increasing “input multiplicity,” a common feature in biological networks where a lone signal influences multiple parts of the network at once.
This suggests that the decision-making power of cells comes not from its size, but from this key factor. Floyd explained this further suggests naturally evolved circuits may have many interconnected components that allow for complex decision-making with fewer parts.
This understanding of the network’s rules and limitations could lead to new technologies.
“Our work aims to understand the computational power of these chemical reaction-based ‘computers’ and identify the features that enable them to perform more complex functions,” said Floyd.
“Since we can now in principle train chemical reaction networks to perform complex classification tasks, we can imagine making ‘smart’ chemical processes.”
These human-designed, “smart” networks could be embedded in biological environments to sense and respond to signals for applications such as medical diagnostics.
The research was supported by the Physics Frontier Center for Living Systems and used the resources of the UChicago Research Computing Center.
Citation: Floyd, C., Dinner, A. R., Murugan, A., & Vaikuntanathan, S. Limits on the computational expressivity of non-equilibrium biophysical processes. Nature Communications, Aug. 5, 2025.
Funding: National Institutes of Health, National Science Foundation, UChicago.