Synaptic Conductance Inference for Oscillating Networks

See the synaptic input behind a rhythm.

SCION is an analysis workflow for recordings from cells embedded in periodically active networks. Give it current, voltage, and a cycle reference collected while you step current or voltage. It returns excitatory and inhibitory conductance profiles over one normalized cycle.

Data channels
4
Phase bins
1000
Slice files
59
Accepted
50

Experimental Contract

Record one cell. Read out the network.

The recorded cell sits inside a periodically active network and receives phase-locked synaptic input. SCION uses the cell as a postsynaptic sensor: perturb its membrane current or voltage, align the responses to the network cycle, and infer the conductances that best explain the observed I-V relationships.

What each exported file contains

1 Time sample time in seconds
2 Current channel injected current in current clamp; holding current in voltage clamp
3 Voltage channel membrane potential in current clamp; command voltage in voltage clamp
4 Cycle reference integrated XII activity here; any reliable phase marker in another preparation

What You Bring

A rhythm, a perturbation axis, and a phase clock.

The method is agnostic to whether the rhythm is spontaneous, pharmacologically maintained, or induced by stimulation. What matters is that repeated cycles can be aligned and that the recording samples enough current-voltage space to estimate a local relationship at each phase.

Postsynaptic recording

Intracellular current and voltage from one identified cell while the surrounding network remains active.

Command spread

Current steps or voltage steps distributed across many cycles, giving the regression something to fit.

Cycle reference

A nerve burst, stimulus trigger, population signal, or other marker that assigns each sample a phase.

Method

SCION treats phase as the independent variable.

The algorithm does not average away the command perturbations. It uses them. For each phase of the cycle, samples collected at different command levels form a small I-V experiment inside the ongoing network rhythm.

Representative current-clamp and voltage-clamp command steps aligned to a respiratory cycle reference.

Step 01

Record across command levels while the network keeps cycling.

Current-clamp and voltage-clamp files provide the same ingredients: a current coordinate, a voltage coordinate, and a cycle reference. SCION uses the full accepted interval rather than treating each step as a separate experiment.

Input
current, voltage, time, phase reference
Output
samples spanning many cycles and command levels

Inference Geometry

At every phase, the data ask for a line.

Samples with the same normalized phase are pooled across cycles and command levels. SCION fits I = GtotV + I0 in each bin, so the changing slope reports total conductance and the changing intercept reports the phase-dependent current offset.

The fitted slope-intercept trajectory forms a wedge. Its envelope estimates the inhibitory boundary for that recording; with the excitatory reversal fixed, the mixed synaptic drive separates into Gexc/gleak and Ginh/gleak.

Geometric conductance inference workflow showing phase-binned I-V regressions, wedge geometry, and reconstructed conductance traces.
One slice recording, shown from phase-binned I-V fits to separated conductance profiles.

Fit Checklist

Use SCION when the rhythm gives you repeatable phase.

The method was developed for respiratory CPG preparations, but its real target is broader: periodically active networks where synaptic input to a recorded cell can be sampled repeatedly at different command levels.

Good fit

  • Cycles repeat often enough to populate phase bins.
  • The network period is much longer than the membrane time constant and spike timescale.
  • The recording has enough current or voltage spread for local I-V regression.
  • A reference signal marks the cycle consistently.
  • The analyzed interval is reasonably stationary.

Use caution

  • Single events or drifting rhythms with no stable phase reference.
  • Recordings with little command spread near key phases.
  • Strong intrinsic nonlinearities dominating the sampled voltage range.
  • Network dynamics changing faster than the membrane can equilibrate.

Respiration As Testbed

Two respiratory preparations demonstrate the same method at different scales.

The point is not that SCION belongs only to respiration. Respiration is a demanding validation arena: the network is rhythmic, the postsynaptic inputs are mixed, and the preparation can be compared across intact and reduced circuit states.

Published method paper

In situ respiratory CPG

Molkov and colleagues introduced the technique in mature rat brainstem-spinal cord recordings, using respiratory motor outputs as phase references to infer excitatory and inhibitory conductance profiles in active respiratory microcircuits.

Read eLife RP101959

This repository

Rhythmic medullary slice

This site exposes the same inference workflow on VgluT2 and VGAT whole-cell recordings from neonatal mouse slices. The dashboard keeps every accepted and rejected recording visible, including parameters, snapshots, cycle counts, and conductance summaries.

Open the audit trail

Transferable idea

Your rhythmic preparation

Replace XII or phrenic activity with your own phase marker: stimulation pulse, population activity, motor output, imaging trace, or another clock. If the cell receives repeatable phase-locked synaptic drive, SCION gives you a way to ask which conductance carries it.

git clone https://github.com/ymolkov/synaptic-inputs-slices
cd synaptic-inputs-slices
make analysis
Open the repository

What This Repo Shows

The slice data are the inspectable worked example.

The generated figures and dashboard are not decorative summaries. They are the audit path from command protocol to single-cell profile to population-level circuit interpretation.

Conductance-weighted inferred circuit diagram for VgluT2 and VGAT inspiratory and expiratory populations.

Biological Readout

For the slice, the conductance profiles imply a compact respiratory circuit.

In this worked example, inspiratory VgluT2 cells carry the excitatory kernel, VGAT expiratory cells provide tonic expiratory inhibition, and VGAT inspiratory cells add phasic inhibition near the inspiratory burst.

  • Excitation is strongest around inspiration in the VgluT2-I population.
  • Inhibition appears both as tonic expiratory drive and phasic inspiratory drive.
  • The slice profile is simpler than the richer in situ respiratory connectome.

Preparation Matters

Same analysis, different preparation, different circuit answer.

The in situ paper uses the method to resolve a richer respiratory CPG organization. The slice study uses the same conductance logic to expose the reduced circuit retained in an isolated preparation. That contrast is the advertisement: SCION is useful because it separates method from preparation.

Comparison of compact in vitro circuit and expanded in situ inhibitory connectome.
Respiratory slice output read alongside the expanded in situ architecture.

Dashboard

The dashboard is the method under a glass cover.

Browse all recording-level outputs: command mode, cycle count, accepted/rejected status, inferred parameters, source files, thumbnails, and full-resolution analysis snapshots.

How to Cite

If SCION informs your work, please cite both papers.

The conductance inference technique was introduced and validated in the eLife method paper. The slice study below applies it to genetically identified preBötC populations.

Method paper

Molkov YI, Borgmann A, Koizumi H, Hama N, Zhang R, Smith JC. Inference technique for the synaptic conductances in rhythmically active networks and application to respiratory central pattern generation circuits. eLife 13:RP101959 (2025).

doi.org/10.7554/eLife.101959

Slice study (this site)

Molkov YI, Koizumi H, Smith JC. Functional Synaptic Interactions and Inhibitory Circuitry of the PreBötzinger Complex in the Rhythmic Slice. bioRxiv (2026).

doi.org/10.64898/2026.05.23.727419