Imagine a mathematical approach to brain research that models the mechanistic basis of brain disorders. Then envision a future in which these models are used to offer personalized medicine to individuals with mental health issues. The field of computational psychiatry, a recent subdiscipline that develops and uses mathematical models of brain function and structure to explain dysfunction in psychiatric disorders, shows great promise in meeting those needs.
Within the Department of Psychiatry’s Division of Neurocognition, Neurocomputation, and Neuro-genetics, or N3 Division, two researchers are spearheading research in collaborative computational psychiatry with the goal of making inferences about the underlying biology of the brain’s neural system. Alan Anticevic, PhD, assistant professor of psychiatry and of psychology, and co-director of N3, leads a clinical neuroimaging laboratory focused on severe psychiatric illness; and John Murray, PhD ’13, assistant professor of psychiatry, of neuroscience, and of physics; directs a laboratory that employs computational modeling to improve understanding of neural function and dysfunction in psychiatric disorders.
Before joining forces, both scientists had observed that many neuroimaging findings, while powerful in mapping the brain, lack a mechanistic interpretation. Moreover, both saw a critical need to translate theoretical findings to clinical research in a principled way. “When we first met, it was obvious to us that a computational-empirical partnership was one of the few ways we could make a practical impact,” Anticevic says.
While neuroimaging has come a long way in the last century, spatial and temporal resolutions are still limited; no human imaging can “see” at the level of individual neurons. “Even when we can experimentally manipulate brain function with pharmacology, it’s difficult to make inferences about what’s happening at the level of neurons,” says Anticevic. “Even a single unit of fMRI measurement, a voxel, has a rich circuit landscape made of millions of neurons and kilometers of axons that we just can’t see.”
To bridge the gap between the seen and unseen, Murray’s team creates mathematical models of the brain’s electrochemical activity at the cellular level. “The models are used to study how perturbations at the mechanistic level can propagate upward to impact brain-wide functioning,” Murray says. “Studying the properties of neurons and synapses allows us to describe those interactions and activity through dynamic equations that resemble equations for resistors and capacitors.” Murray’s team then builds models that incorporate some of the details that simulate neural activity, which are later evaluated with neuroimaging experiments.
“There are many ideas about what may happen at the neuronal level that drives the cascade of alterations that produce abnormal behavior,” says Anticevic. “For instance, one prominent hypothesis in psychosis research is that there may be something ‘off’ in the way the NMDA receptor is working.”
The NMDA receptor, or NMDAR, is a protein molecule that lies within almost all brain-cell membranes and mediates synaptic interactions. Anticevic describes it as a tiny logical gate with two requirements for excitation: first, specific molecules must bind to it; second, it must receive a signal from another cell. When these events co-occur, the channel opens to allow the flow of positively charged ions, thus changing the cell’s polarization. “With psychosis, we hypothesize that something goes quite wrong with this core computational machinery,” he says. Along with others in the field who are testing this hypothesis, John Krystal, MD ’84, HS ’88, the Robert L. McNeil, Jr. Professor of Translational Research, professor of psychiatry and of neuroscience; and chair of the Department of Psychiatry, observes that giving healthy individuals low doses of ketamine temporarily blocks a specific kind of NMDA receptor found in cells that calm the brain. The result is a transient state with some of the behavioral symptoms associated with psychosis or schizophrenia.
Anticevic’s research includes administering ketamine in order to study the immediate effects of the drug that induce psychosis. “We observe pharmacological effects with imaging and behaviorally,” he says. As the behavioral effects of ketamine occur, specific changes in neural activity across the brain are captured by fMRI. “With evidence from experiments, we can put a hypothesis into a mathematical framework that predicts what would happen if we block NMDA receptors everywhere equally or somewhere more preferentially,” says Anticevic.
“Only recently has modeling been able to make experimentally testable predictions for neuroimaging,” Murray says. “Being able to design experiments and analyses around models is critical because we need these particular types of data and analyses to iteratively test and refine the models and improve them.”
Computational psychiatry hasn’t yet crossed the bridge into clinical practice. However, the researchers hope that their work will reveal the diverse mechanisms of brain disorders and provide models that can be used to make predictions at the level of the individual patient. These models in turn will minimize clinical guesswork regarding causes and treatments. The National Institute of Mental Health (NIMH) actively supports research in the field of computational neuroscience.
Anticevic noted that collaborative research within the N3 framework offers a multidisciplinary approach that attracts researchers from physics, mathematics, computer science, computer engineering, neuroscience, psychiatry, clinical psychology, and cognitive neuroscience. “The problem doesn’t care about your background,” says Anticevic. “There isn’t any one skill set that can tackle this challenge. It’s bigger than any one lab.”