Advancing Computational Neurobiology

One powerful application of multi-scale modeling is the ability to identify mechanistic changes at the cellular scale that give rise to emergent network, circuit, and behavioral effects at the clinical scale. We have worked on this problem in the context of neurometabolism, systematically testing how modulation of TCA-cycle inputs and neuronal insulin resistance affect brain function—from single-neuron firing dynamics to axon conduction velocity to whole-brain network behavior to cognition.

In collaboration with Doug Rothman, PhD (Yale University), we currently have a manuscript under review that integrates metabolic flux equations for neurotransmitter cycling, accurately predicting glutamate and GABA dynamics in response to ketone and glucose bolus administration described in our grant’s Acute Metabolic Challenge Test. This work forms the basis for the Neuroblox Metabolic Module, a plug-in to Neuroblox Circuits (which includes our corticostriatal circuit currently recently published in Nature Communications).

Complementary to these efforts, we developed unbiased data-driven methods for reducing control circuit models (including multi-scale models) to their key components. These make it possible to determine the inherent limits on what each scale and modality can and can’t measure.

These findings provide the first comprehensive characterization of how metabolic state directly influences neural synchrony and connectivity in the aging brain, offering a mechanistic foundation for therapeutic approaches targeting age-related cognitive decline. The use of highly controlled metabolic interventions that can be translated between human neuroimaging (fMRI, MRS) and neurons (patch clamp, field recordings) has allowed us to make significant methodological advances in our understanding of neurometabolism across the lifespan.

A major area of research by our lab has been neurometabolism and its predictive and preventative impact on the clinical trajectory of brain aging. In particular, we have worked at all scales to systematically develop a mechanistic understanding of how metabolic interventions can restore age-related disruptions in brain network function.

Our research demonstrates that ketosis—achieved through ketone administration—strengthens brain-wide signaling by modulating key neurobiological processes: regulating K+ ion channels that are disrupted by aging, downregulating inhibitory GABA neurotransmission while optimizing excitatory glutamate signaling, and elevating antioxidants alongside energy metabolism markers.

Our team has pioneered methods to identify and quantify control circuit regulation of intact feedback loops from human neuroimaging data. This matters because distinct types of neurotransmitter-based dysregulation for key circuits (prefrontal-limbic for threat, corticostriatal for learning and reward, ventral stream and thalamocortical loop for sensory integration) are thought to underlie most disorders of cognition, mood, and behavior.

However, “dysregulation” cannot be characterized by standard neuroimaging analyses focused solely on activation maps and/or connectivity. Our contributions include adapting techniques from complex systems analysis and control systems engineering to neuroimaging time series, and quantifying circuit-based regulatory parameters such as feedback strength, sensitivity to perturbation, and control error.

A complementary body of work establishes critical methodological foundations for translating neuroimaging research into reliable clinical neurodiagnostic tools. This work addresses challenges in extracting robust individual-specific parameters from noisy neural data.

Our research provides solutions to measurement noise contamination through data-driven correction methods for scanner-induced distortions in fMRI time series and principled parameter estimation techniques for stochastic neural processes. A key contribution is the development of computationally efficient modeling frameworks that significantly increase the speed and flexibility of dynamic causal modeling—essential for clinical applications requiring real-time parameter estimation.

The integration of novel behavioral paradigms with advanced computational methods enables functional dissection of neural circuit activity at the individual level, moving beyond group-averaged approaches toward personalized neurodiagnostics. This work directly addresses the gap between research-grade neuroimaging and clinical implementation.

The first article introduces a dynamic phantom developed by our team, BrainDancer™ (Patent 11,061,093), which amplifies signal-to-noise ratio to enable single-subject-level fMRI.