A team of researchers from Dartmouth College, MIT, and the State University of New York at Stony Brook has developed a new computational brain model that closely mirrors biological and physiological processes. The model successfully learned a simple visual category learning task with results that matched those observed in laboratory animals. According to the scientists, the model achieved this without being trained on animal data, instead relying on fundamental principles of neural connectivity and communication.
"It's just producing new simulated plots of brain activity that then only afterward are being compared to the lab animals. The fact that they match up as strikingly as they do is kind of shocking," said Richard Granger, professor of Psychological and Brain Sciences at Dartmouth and senior author of the study published in Nature Communications.
The research aims not only to advance understanding of normal brain function but also to explore differences in disease states and potential interventions. Earl K. Miller, co-author and Picower Professor at MIT's Picower Institute for Learning and Memory, noted that he, Granger, and others have founded Neuroblox.ai to develop biotech applications based on these models. Lilianne R. Mujica-Parodi, a biomedical engineering professor at Stony Brook who serves as Lead Principal Investigator for the Neuroblox Project, is CEO of the company.
Dartmouth postdoctoral researcher Anand Pathak created the model. Unlike many previous models that focus either on detailed neuron-to-neuron connections or broader architectural features alone, this one integrates both levels. Pathak explained: "We didn't want to lose the tree, and we didn't want to lose the forest."
In technical terms, small circuits known as "primitives" mimic real neuronal electrical and chemical interactions for basic computations within a modeled cortex. For example, excitatory neurons receive input through synapses influenced by glutamate before interacting with inhibitory neurons in a competitive pattern seen in actual brains.
On a larger scale, four regions—cortex, brainstem, striatum, and tonically active neuron (TAN) structures—work together for learning tasks. During training sessions where patterns of dots had to be categorized into broader groups, TANs introduced variability through acetylcholine bursts so that different actions could be explored before consistent performance emerged via strengthened cortical-striatal connections.
As learning progressed in both animals and simulations alike, increased synchrony between cortex and striatum was observed in beta frequency rhythms during correct judgments—a phenomenon documented previously by Miller’s research.
Unexpectedly, about 20 percent of modeled neurons showed activity predictive of errors rather than successes. When these “incongruent” neurons influenced circuits during categorization tasks, incorrect choices resulted. Upon reviewing existing animal data from Miller’s lab with this insight from modeling work in mind, researchers found similar patterns previously overlooked.
"Only then did we go back to the data we already had, sure that this couldn't be in there because somebody would have said something about it, but it was in there and it just had never been noticed or analyzed," Granger said.
Miller suggested such cells may help adapt behavior when rules change: "It's all well and good to learn the rules of a task but what if the rules change? Trying out alternatives from time to time can enable a brain to stumble upon a newly emerging set of conditions." Recent studies at another Picower Institute lab support this idea across humans and other animals.
The team is now expanding their model’s capabilities by adding more brain regions and additional neuromodulatory chemicals while testing how pharmaceutical interventions affect its dynamics.
Other contributors include Scott Brincat, Haris Organtzidis, Helmut Strey, and Evan Antzoulatos. Funding was provided by organizations such as The Baszucki Brain Research Fund (United States), Office of Naval Research (ONR), and Freedom Together Foundation.
