Imagine this: Scientists have created a brain model so accurate that it not only mimics how animals learn but also reveals hidden neural activity previously unnoticed in animal studies. Fascinating, right? This groundbreaking research, conducted by a team from Dartmouth College, MIT, and the State University of New York at Stony Brook, is revolutionizing our understanding of the brain. But here's where it gets really interesting…
This new computational model, built to mirror the brain's biology and physiology, achieved remarkable results without any prior training on animal experiment data. It was meticulously designed from scratch to replicate how neurons connect, communicate, and influence cognition and behavior. When tasked with a simple visual category learning task (identifying patterns of dots), the model's neural activity and behavioral outcomes closely mirrored those of lab animals, even exhibiting the same erratic learning progress.
"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," says Richard Granger, a professor at Dartmouth and senior author of the study published in Nature Communications.
The primary goal of this model, and its subsequent iterations, is not only to understand the brain's inner workings but also to explore how it functions differently in disease and how interventions might correct these issues. Co-author Earl K. Miller, a professor at MIT, and others have launched a company, Neuroblox.ai, to develop the model's biotech applications. Their aim is to create a platform for biomimetic brain modeling to improve the discovery, development, and testing of neurotherapeutics. For instance, drug development and efficacy testing could potentially happen earlier in the process, reducing the risk and cost of clinical trials.
So, how does it work? The model, created by Dartmouth postdoc Anand Pathak, incorporates both microscopic details, such as how individual neurons connect, and large-scale architecture, including the influence of neuromodulatory chemicals like acetylcholine. The team ensured their designs adhered to real-brain constraints, like how neurons synchronize. Many existing models focus on either the small or large scale, but not both.
Pathak explained, "We didn't want to lose the tree, and we didn't want to lose the forest."
The "trees," or "primitives" in the study, are small circuits of neurons that connect based on the electrical and chemical principles of real cells. These circuits perform fundamental computational functions. For example, within the model's cortex, excitatory neurons receive input from the visual system via connections affected by the neurotransmitter glutamate. These excitatory neurons then connect with inhibitory neurons in a competition, regulating information processing.
At a larger scale, the model includes four brain regions essential for learning and memory: the cortex, brainstem, striatum, and a "tonically active neuron" (TAN) structure. The TAN injects "noise" into the system via bursts of acetylcholine. When the model categorizes dot patterns, the TAN initially introduces variability, enabling the model to learn through varied actions and outcomes. As learning progressed, the cortex and striatum circuits strengthened, suppressing the TAN and allowing the model to act with increasing consistency.
As the model engaged in the learning task, it displayed real-world properties, including a dynamic commonly observed by Miller in his research with animals. As learning progressed, the cortex and striatum became more synchronized in the "beta" frequency band of brain rhythms, and this increased synchrony correlated with times when the model (and the animals) made the correct category judgement about what they were seeing.
But here's where it gets controversial... The model revealed a group of neurons, about 20%, whose activity predicted errors. These "incongruent" neurons, when influencing circuits, led the model to make incorrect category judgements. Initially, the team thought it was a model quirk. However, upon reviewing data from animal experiments, they discovered these neurons were present but had been overlooked.
Miller suggests these counterintuitive cells might serve a crucial purpose: enabling the brain to adapt to changing rules. Trying out alternatives can help the brain discover new conditions.
While the model exceeded expectations, the team continues to expand its capabilities. They are adding more regions, new neuromodulatory chemicals, and testing the effects of interventions like drugs.
In addition to Granger, Miller, Pathak, and Mujica-Parodi, the paper's other authors are Scott Brincat, Haris Organtzidis, Helmut Strey, and Evan Antzoulatos.
The research was supported by The Baszucki Brain Research Fund, United States, the Office of Naval Research, and the Freedom Together Foundation.
What are your thoughts? Do you find the discovery of these "incongruent" neurons surprising? How might this research impact the future of neurotherapeutics? Share your opinions in the comments below!
