Circuit decoupling with surrogate compensation for baseline tuning in biophysically detailed neural models

The baseline regime of a neural circuit shapes its dynamical behavior. Establishing an appropriate baseline is therefore an important part of tuning a computational circuit model. This baseline can be specified in terms of target population firing rates and achieved by adjusting the background inputs to each population. Due to recurrent coupling, these inputs need to be tuned jointly, requiring a high-dimensional search that is especially costly for biophysically detailed circuit models. This remains a major bottleneck since current optimization methods are limited in their ability to tackle this problem. Decoupling the circuit by removing all the connections between its cells allows background inputs to be tuned independently, but the solution may not hold when the connectivity is restored. We developed a compensated decoupling method in which missing recurrent inputs are replaced by surrogate inputs from spike generators firing at the corresponding target rates. The resulting input approximates the recurrent drive expected at the target regime, so the background inputs tuned in this setup may remain valid after restoring connectivity. We applied the method to a 43-population biophysically detailed model of the macaque auditory thalamocortical circuit, which had previously proved challenging to fit closely to experimental data. First, we tuned the background input independently for each population in the fully decoupled setup. Next, we tested the solution in the isolated cortical and thalamic subcircuits, preserving recurrent connections within each subcircuit. The cortical subcircuit had two coexisting regimes and settled into the off-target one by default. A gradual transition from surrogate to recurrent self-excitation allowed it to reach the near-target regime. The thalamic subcircuit exhibited a single regime deviating from the target due to imperfect surrogate compensation. Jointly tuning a separate tonic input correction for each thalamic population substantially reduced the errors. Finally, we tested the solution in the full thalamocortical circuit, combining the tuned background inputs, thalamic corrections, and gradual restoration of self-excitation. The full circuit consistently reached a near-target regime across multiple random model realizations. The approach yielded an improved macaque auditory thalamocortical baseline model, providing a community resource for investigating the auditory thalamocortical system's dynamics. Together, the results demonstrate that our method can facilitate baseline tuning in a biophysically detailed model by enabling decoupled optimization and guiding subsequent refinement.

Authors

Institutions

Publication Details

Journal
bioRxiv (Cold Spring Harbor Laboratory)
Published
2026-11-16
DOI
https://doi.org/10.64898/2026.11.13.766494
Primary Topic
Neural dynamics and brain function
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Circuit decoupling with surrogate compensation for baseline tuning in biophysically detailed neural models

Новиков Никита Александрович, Samuel A. Neymotin, Salvador Durá-Bernal, James Chen et al.
bioRxiv (Cold Spring Harbor Laboratory)
Neural dynamics and brain function
preprint

Circuit decoupling with surrogate compensation for baseline tuning in biophysically detailed neural models

Новиков Никита Александрович, Samuel A. Neymotin, Salvador Durá-Bernal, James Chen, Ethan Irby, Scott McElroy
preprint en

Abstract

The baseline regime of a neural circuit shapes its dynamical behavior. Establishing an appropriate baseline is therefore an important part of tuning a computational circuit model. This baseline can be specified in terms of target population firing rates and achieved by adjusting the background inputs to each population. Due to recurrent coupling, these inputs need to be tuned jointly, requiring a high-dimensional search that is especially costly for biophysically detailed circuit models. This remains a major bottleneck since current optimization methods are limited in their ability to tackle this problem. Decoupling the circuit by removing all the connections between its cells allows background inputs to be tuned independently, but the solution may not hold when the connectivity is restored. We developed a compensated decoupling method in which missing recurrent inputs are replaced by surrogate inputs from spike generators firing at the corresponding target rates. The resulting input approximates the recurrent drive expected at the target regime, so the background inputs tuned in this setup may remain valid after restoring connectivity. We applied the method to a 43-population biophysically detailed model of the macaque auditory thalamocortical circuit, which had previously proved challenging to fit closely to experimental data. First, we tuned the background input independently for each population in the fully decoupled setup. Next, we tested the solution in the isolated cortical and thalamic subcircuits, preserving recurrent connections within each subcircuit. The cortical subcircuit had two coexisting regimes and settled into the off-target one by default. A gradual transition from surrogate to recurrent self-excitation allowed it to reach the near-target regime. The thalamic subcircuit exhibited a single regime deviating from the target due to imperfect surrogate compensation. Jointly tuning a separate tonic input correction for each thalamic population substantially reduced the errors. Finally, we tested the solution in the full thalamocortical circuit, combining the tuned background inputs, thalamic corrections, and gradual restoration of self-excitation. The full circuit consistently reached a near-target regime across multiple random model realizations. The approach yielded an improved macaque auditory thalamocortical baseline model, providing a community resource for investigating the auditory thalamocortical system's dynamics. Together, the results demonstrate that our method can facilitate baseline tuning in a biophysically detailed model by enabling decoupled optimization and guiding subsequent refinement.

bioRxiv (Cold Spring Harbor Laboratory)
Nathan Kline Institute for Psychiatric Research (US), SUNY Downstate Health Sciences University (US)
New York State Department of Health, National Institutes of Health
Neural dynamics and brain function
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon