Summary: An artificial neural network has revealed a plausible mechanism behind impaired decision-making often observed in schizophrenia: reduced activity of NMDA receptors in cortical circuits.
Source: eLife
Researchers have developed a computer ‘brain circuit’—a spiking artificial neural network—that reproduces human-like decision-making and pinpoints how synaptic changes may produce the cognitive deficits seen in psychiatric disorders, according to a new study published in eLife.
The model suggests a specific mechanism for impaired evidence integration in schizophrenia: diminished function of NMDA receptors at synapses. These findings clarify how synaptic alterations can alter circuit dynamics and behaviour, and may guide future therapeutic strategies for neuropsychiatric conditions.
“A core obstacle in psychiatry is connecting changes at synapses to the cognitive processes that underlie disorders like schizophrenia,” says lead author Dr. Sean Cavanagh, an MBPhD student at the UCL Queen Square Institute of Neurology, London. “Computational circuit models let us translate synaptic perturbations into predictions about neural activity and behaviour, which can then be tested experimentally.”
The research team implemented a neural circuit that mimics how the brain accumulates evidence and reaches a decision. They focused on decisions that require integrating multiple pieces of information—for instance choosing a holiday destination requires weighing cost, weather and cultural factors into a single choice.
First, the researchers verified that their model reproduced a known human decision bias called the pro-variance bias. This bias reflects a tendency to favor options with more variable evidence: when comparing two choices, people often prefer one that is outstanding on a single attribute but weak on others over an option that is consistently mediocre across attributes.
To ground the modelling in data, the team trained two macaque monkeys on a decision task and recorded their behaviour. On each trial the animals viewed two sequences of eight vertical bars, one sequence on the left and one on the right, and had to decide which side had the taller average height.
Across nearly 30,000 trials, the monkeys’ choices showed the same pro-variance bias observed in humans: they tended to prefer the option with greater variability across evidence samples. The researchers examined how the timing and variance of evidence influenced those choices to better constrain their circuit model.
The computational circuit includes two populations of excitatory neurons—one representing the left option, the other the right—and an interconnected population of inhibitory neurons that stabilises network activity. High activity in one excitatory group signals a decision for its corresponding option. When driven with the same task inputs used with the monkeys, the circuit reproduced the pro-variance bias, identifying a plausible neural mechanism for the behaviour.
To explore how this decision process could be disrupted in neuropsychiatric disorders, the team modelled reduced NMDA receptor function selectively on excitatory or inhibitory neurons. NMDA receptors shape synaptic integration and excitation/inhibition balance; changes to their function shifted the model’s decision performance depending on where the hypofunction occurred.
Ketamine, an NMDA receptor antagonist often used as an experimental model of schizophrenia, transiently produces many schizophrenia-like symptoms in healthy people. The researchers administered ketamine to the monkeys and re-tested their decision-making behaviour.
Following ketamine, the monkeys’ decision accuracy declined, but importantly the pro-variance bias remained and in some analyses was increased. In the model, these behavioural changes were consistent with a shift toward lower cortical excitation/inhibition balance caused by NMDA receptor hypofunction predominantly affecting excitatory neurons.
Taken together, the experimental and modelling results indicate that reduced NMDA receptor activity at excitatory synapses can produce the kinds of biased and degraded decision-making seen in schizophrenia and related disorders. By linking synaptic mechanisms to circuit dynamics and behaviour, the study provides a mechanistic bridge across levels of explanation that could help direct future intervention strategies.
“Decision-making biases can be part of normal behaviour, or they can reflect underlying deficits in neuropsychiatric conditions,” notes senior author Professor Steve Kennerley, Professor of Cognitive Neuroscience at the UCL Queen Square Institute of Neurology. “Building mechanistic accounts of brain disorders requires tightly controlled behavioural tasks, comparisons between typical and perturbed states, and computational models that span synapses to behaviour. This work was possible through collaborations with Oxford and Yale, and it offers a circuit-level mechanism linking synaptic NMDA receptor hypofunction to decision-making abnormalities.”
About this auditory neuroscience research article
Source:
Garvan Institute of Medical Research
Contacts:
Emily Packer – eLife
Image Source:
The image is in the public domain.
Original Research: Open access
“A circuit mechanism for decision-making biases and NMDA receptor hypofunction” by Sean Edward Cavanagh (corresponding author), Norman H. Lam, John D. Murray, Laurence Tudor Hunt, Steven W. Kennerley. eLife. DOI: 10.7554/eLife.53664
Abstract
A circuit mechanism for decision-making biases and NMDA receptor hypofunction
Decision-making biases can reflect normal strategy or be symptomatic of neuropsychiatric disorders. Using behavioural psychophysics, spiking-circuit modelling and pharmacological manipulation, the authors examined how evidence integration generates decision biases. Monkeys displayed a pro-variance bias (PVB): a tendency to choose options with more variable evidence. A spiking circuit model reproduced this bias, suggesting a neural basis. Simulating NMDA receptor hypofunction on excitatory versus inhibitory neurons produced distinct effects in the model; these predictions were tested experimentally with ketamine, an NMDA receptor antagonist and pharmacological model of schizophrenia. Ketamine increased the PVB and reduced decision accuracy, consistent with lowered cortical excitation/inhibition balance arising mainly from NMDA receptor hypofunction on excitatory cells. These results offer a circuit-level account linking synaptic changes to behavioural decision-making biases relevant to neuropsychiatric disorders.