Neuroblox Biomimetic Model of Corticostriatal Micro-assemblies Discovers New Neural Code
Poster No:
1120
Submission Type:
Abstract Submission
Authors:
Lilianne Mujica-Parodi1, Richard Granger2, Earl Miller3, Helmut Strey1, Christopher Rackauckas3, Alan Edelman3, Anand Pathak2, Haris Organtzidis1, Mason Protter1
Institutions:
1State University of New York at Stony Brook, Stony Brook, NY, 2Dartmouth College, Hanover, NH, 3Massachusetts Institute of Technology, Cambridge, MA
First Author:
Co-Author(s):
Introduction:
The Neuroblox software platform is designed for computational neuroscience and psychiatry applications, with the goal of simulating disorders and interventions that affect perceptual, cognitive, emotional, and motor functions. Thus, Neuroblox circuits are composed of a common library of biomimetic computational primitives (BCPs) that underlie those functions. Just as an electrical circuit can be composed of signal processing elements such as filters, gains, resistors, capacitors, etc., the brain uses spiking-neuron-based micro-circuits that generate its most fundamental "processing" functions. These are the smallest-scale "blocks" in Neuroblox. BCPs can be combined to assemble micro-circuits, which in turn can be combined to generate macro-circuits. Once these composite blocks are defined, they, like BCPs, can be used repeatedly as needed. This hierarchical "blocks of blocks" approach allows Neuroblox to scale upwards: micro to macro. The fact that BCPs are biomimetic is crucial because an intervention's effects on underlying mechanisms at the BCP scale will alter basic neural processing, with emergent effects at the macro-circuit and clinical neuroimaging/symptom scales.
Methods:
Our tools range from control circuit system identification to brain circuit simulations bridging scales from spiking neurons to fMRI-derived circuits, parameter-fitting models to neuroimaging data, interactions between the brain and other physiological systems, experimental optimization, and scientific machine learning. Neuroblox.jl is based on a library of modular computational building blocks ("blox") in the form of systems of symbolic dynamic differential equations that can be combined to describe large-scale brain dynamics. Once a model is built, it can be simulated efficiently and fit electrophysiological and neuroimaging data. Moreover, the circuit behavior of multiple model variants can be investigated to aid in distinguishing between competing hypotheses.
Results:
We first integrated data from the spiking-neuron-scale literature to extract a subset of well-validated BCPs: Superficial Cortical (SCORT), Matrisomal MSN (MMSN), Striosomal MSN (SMSN), Ascending Type 1 (ASC1), Superficial Glu−GABA−DA, and Cholinergic Tonically Active Neurons (CTAN). These BCPs are the smallest-scale "blocks" in Neuroblox. BCPs can be combined to assemble micro-circuits, which in turn can be combined to generate macro-circuits. We then integrated data from the composite circuit-scale literature to define each subcircuit: Cortex Subcircuit, Striatum Subcircuit, TAN Subcircuit, Brainstem Subcircuit, Thalamus Subcircuit. This hierarchical "blocks of blocks" approach allows Neuroblox.jl to scale upwards: micro to macro. In this manner, using Neuroblox.jl we integrated very different types of data and models from the scientific literature to construct a telencephalic architecture comprising corticostriatal loops, pallidal and thalamic components of the loop, as well as ascending systems with multiple modulatory receptor types and their connectivity. The resulting system produces electrophysiological and behavioral outputs validated by independent data sources, data of a fundamentally different type than that used to create the model. Although each component of the corticostriatal circuit was grounded in spiking-neuron-elicited BCPs, these outputs include complex category learning and state-dependent modulation of corticostriatal beta-frequency synchrony during the stimulus presentation and during the no-stimulus delay working memory (working memory) periods, as well as discovery of new neuronal behavior (Pathak, 2024).
·Multiscale Corticostriatal Model: (a) BCP microassemblies (b) generate fundamental processing functions, (c) which then are assembled into composite circuits.
·Experimental Validation: (a) Behavioral learning over trials, (b-c) synaptic strength/activity across trials, (d) corticostriatal phase locking for simulation vs. macaque electrophysiology/behavior
Conclusions:
To demonstrate proof of concept, we integrated knowledge from the scientific literature to develop a novel corticostriatal model, which was then validated on macaque electrophysiology (spiking-neuron, LFP) and behavioral data. Importantly, our model not only matched known results but "discovered" previously unknown neuronal behaviors that were also empirically validated.
Higher Cognitive Functions:
Higher Cognitive Functions Other 2
Learning and Memory:
Learning and Memory Other
Modeling and Analysis Methods:
Classification and Predictive Modeling 1
Methods Development
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Microcircuitry and Modules
Keywords:
Cognition
Computational Neuroscience
ELECTROPHYSIOLOGY
Learning
Modeling
1|2Indicates the priority used for review