Introduction
Every so often, a simulation surprises you. Not because of a coding error — but because it behaves too much like life.
During a recent validation run inside BioSynapStudio, we observed something unexpected.
Three neurons — all built from the same Hodgkin–Huxley core, running on identical solver paths — each produced strikingly different firing behaviours:
- A bursting pattern under depolarised leak and reduced potassium conductance,
- A post-inhibitory rebound spike after release from inhibition, and
- A near-threshold doublet, where the neuron fired twice from a barely sufficient depolarisation.
None of these modes were pre-programmed or hard-coded. They emerged from tuning biophysical parameters, not from changes in algorithm or architecture.
This is significant because it reflects what happens in biology: neurons with the same underlying physics can behave in entirely different ways depending on their ionic environment and internal balance. When that same principle holds in a synthetic system, it suggests the simulation is capturing more than equations — it’s capturing the logic of life itself.
What Happened
The observation arose during a standard validation cycle in BioSynapStudio, using a point-neuron Hodgkin–Huxley model built around three canonical ionic conductances: sodium (Na⁺), potassium (K⁺), and passive leak.
All tests were run through the same solver path and time-step precision, ensuring that only biophysical parameters, not code paths, determined the outcome.
In the baseline configuration, the neuron rested near –65 mV with no injected current — a stable, quiet state exactly as expected.
But when the leak potential was slightly depolarised and potassium conductance (gK) reduced, a subtle oscillation began to form in the membrane potential trace.
Instead of decaying, these oscillations amplified until the cell entered a sequence of rhythmic, self-sustaining spikes.
The behaviour matched what electrophysiologists describe as epileptiform bursting: rapid depolarisation–repolarisation cycles without external stimulation, emerging from an ionic imbalance that leaves sodium currents dominant.
No numerical artefacts or code divergences were detected — the same solver produced clean, stable results in control runs under standard conductance values.
To cross-check, two additional protocols were run using identical solver and morphology settings:
- a near-threshold double spike, where minimal depolarisation triggered a paired action potential, and
- a post-inhibitory rebound, where transient inhibition was followed by a single restorative spike.
Each experiment confirmed that the observed behaviour was not a random artefact but a context-dependent emergent response of the same model to biophysically meaningful parameter shifts.
“Each of these traces comes from the same BioSynapStudio core, using identical solver paths; yet by simply tuning biophysical parameters, we can reproduce three hallmark neuronal behaviours: bursting, rebound, and near-threshold doublet spiking.” — Desmond K. Atkins, Qognetix
Figure 1: Membrane Potential Comparisons
These runs collectively demonstrate that BioSynapStudio’s numerical engine can express a continuum of behaviours — from stability to pathological hyperexcitability — all from the same underlying substrate.
Why This Matters
In most neuron simulators, you tell the system what kind of behaviour you want to see. You hard-code the parameters for “bursting,” adjust channel kinetics to induce instability, or inject patterned current pulses to force spiking. It’s a top-down process — behaviour scripted into existence.
BioSynapStudio works differently.
All three outcomes described above — the double spike, the rebound, and the burst — emerged from the same mathematical substrate, running through the same solver, with only the biophysical parameters changed. No bespoke logic or code branches were added to define these firing modes; they arose from the natural dynamics of the Hodgkin–Huxley equations expressed with high numerical precision.
This is more than a convenience — it’s a validation of biophysical realism. By simply lowering potassium conductance or adjusting the leak potential, the system transitioned from a quiet resting cell to one exhibiting rhythmic hyperexcitability. That mirrors what happens in real biological tissue: the same neuron can behave very differently depending on its ionic environment.
Traditional engines such as NEURON, NEST, or Brian2 can reproduce these behaviours too, but usually only when the researcher deliberately configures them.
What stands out in BioSynapStudio is that the transitions occur as natural consequences of the model’s conductance ratios, not as scripted presets.
This indicates that the numerical fidelity of the solver preserves subtle instabilities rather than smoothing them away for computational efficiency.
Figure 2: Ionic Current & Conductance Dynamics
Sodium activation produces a rapid inward current followed by a delayed potassium-mediated outward current. The interaction between these opposing forces shapes each spike and drives the transition toward hyperexcitability under altered leak and gK conditions.
That’s important because it moves the system closer to the real physics of excitable membranes. In biology, stability is never absolute — it’s a balancing act between competing currents, constantly adjusting to maintain order. By preserving that delicate balance, BioSynapStudio allows the same cell model to occupy the full spectrum of neuronal behaviour — from silence to seizure — with nothing more than a shift in ionic conductance.
When a digital neuron can do that, it’s no longer just solving equations. It’s beginning to participate in the same physics that biology does.
The Science Behind It
At the microscopic level, the story of neuronal excitability is a story of balance — a perpetual negotiation between inward and outward ionic currents. In the Hodgkin–Huxley framework, sodium (Na⁺) currents drive the cell toward depolarisation, while potassium (K⁺) currents pull it back toward rest. Every spike, pause, and oscillation is the product of how these two forces trade dominance in time.
In a healthy neuron, that balance is exquisitely maintained. When a brief depolarising input arrives, voltage-gated sodium channels open, allowing Na⁺ to rush in and trigger an action potential. Almost instantly, potassium channels open to repolarise the membrane, closing the loop and restoring stability. The result is a single, well-formed spike — a neuron doing exactly what it should.
But when that equilibrium shifts — even slightly — the system’s feedback can change sign. If potassium activation slows, or sodium recovery accelerates, the usual damped oscillation can become positive feedback: each voltage swing amplifies the next. The cell begins to cycle through rapid depolarisation–repolarisation events, entering a self-sustaining loop of activity.
Figure 3: Gating & Phase-Plane Mechanics
m, h, and n gating dynamics illustrate how sodium activation and potassium recovery shape membrane responses. Phase-plane trajectories reveal distinct regimes: a compact loop for doublet spiking, a large limit-cycle orbit for bursting, and a transient excursion for post-inhibitory rebound.
In the BioSynapStudio tests, this regime appeared when the leak potential was biased more depolarised and potassium conductance (gK) was reduced. Those two adjustments tilt the current balance toward inward flow, leaving sodium transiently dominant. Mathematically, the system moves into a limit-cycle instability, where voltage and gating variables chase each other around a closed trajectory instead of settling to rest. Electrophysiologists would recognise this pattern as the core mechanism behind epileptiform bursting.
Importantly, the model did not need exotic equations or stochastic noise to produce it. The same Hodgkin–Huxley kinetics used since 1952, expressed through BioSynapStudio’s precision solver, were sufficient. That fidelity means the engine respects the real physics of excitable membranes — capturing how small parameter shifts can push a neuron from equilibrium into chaos and back again.
Understanding that transition matters.
It turns the simulator from a tool that imitates neurons into one that can investigate them — revealing how normal function degrades into dysfunction, and how the same underlying rules can generate both.
Figure 4: I–V Relationships
Doublet, bursting, and rebound regimes produce distinct I–V loops. The bursting condition shows pronounced hysteresis and a net inward-current region, indicating a shift into a self-sustaining, unstable trajectory. Doublet and rebound traces remain largely monotonic and return to rest.
Implications
The ability to reproduce physiological and pathophysiological behaviour within the same model is more than a technical achievement — it’s a sign of biophysical completeness.
When a synthetic neuron shifts from stability to instability under realistic ionic adjustments, it confirms that the model’s internal dynamics are governed by the same physical relationships that shape living membranes.
🧬 For Neuroscience
For experimental neuroscientists, this matters because it allows a single synthetic framework to map transitions between normal and pathological states without rewriting equations or hard-coding behaviours. By simply altering conductances — lowering potassium, biasing the leak potential, or modifying channel kinetics — the same model can traverse the boundary between healthy spiking and epileptiform bursting.
This offers a practical tool for exploring channelopathies and neuronal excitability disorders, where disease arises not from new mechanisms but from subtle imbalances in existing ones.
Such in silico experiments can complement wet-lab work by providing a high-resolution, repeatable way to probe how conductance ratios, membrane potentials, or temperature factors push a cell toward instability.
🤖 For Synthetic Intelligence
For Qognetix, the implications reach further.
The same substrate that supports realistic pathology also supports realistic cognition. If a synthetic neuron can drift into hyperexcitability for the same reasons a biological one does, it implies that the computational fabric of the system obeys biophysical constraints — not abstract approximations.
That fidelity is foundational to the company’s vision of Synthetic Intelligence: a form of computation that doesn’t just simulate thought but emerges from the same physics that enables it in nature. In BioSynapStudio, no code paths are added to simulate learning, memory, or instability — they appear when the underlying equations are pushed into new regions of parameter space. The bursting seen here isn’t programmed; it’s permitted by the substrate.
💊 For Medicine and Modelling
The same principle applies in biomedical modelling. If a virtual neuron responds to conductance changes in the same way as a biological one, then drug developers, clinicians, and engineers can use it to predict responses to interventions, pharmacological agents, or implant stimulation patterns. It creates a bridge between simulation and experiment, where each can validate the other.
In this way, the very behaviour that in biology signals dysfunction becomes, in simulation, a marker of fidelity and usefulness. By reproducing both normal and pathological dynamics within the same unified model, BioSynapStudio helps close the gap between theoretical neuroscience, medical research, and the long-term pursuit of physics-based synthetic cognition.
Looking Ahead
Every unexpected behaviour in a neural model deserves careful scrutiny — not because it’s a flaw, but because it reveals something about the system’s internal truth. The recent BioSynapStudio results mark a starting point for a deeper validation programme, one designed to confirm that emergent hyperexcitability arises from biophysical principles, not implementation artefacts.
🔬 Validation & Replication
The immediate goal is to replicate these effects systematically across different conditions and configurations.
Upcoming tests will explore:
- Cell morphologies — comparing point neurons with multi-compartment geometries to confirm whether bursting and rebound behaviours scale with spatial complexity.
- Solver parameters — varying integration time steps and numerical precision to identify thresholds at which micro-instabilities appear or disappear.
- Ionic parameter space — sweeping conductance ratios and reversal potentials to map the boundary between stable and unstable regimes.
- Temperature effects — introducing Q₁₀ scaling to examine how physiological temperature shifts influence hyperexcitability.
These experiments will determine whether the observed transitions are general properties of the model’s physics or confined to specific parameter ranges.
To ensure these results are not only internally consistent but externally verifiable, we are now inviting interested partners to collaborate on independent validation and replication of these behaviours. The datasets shown here are derived from controlled single-cell Hodgkin–Huxley runs inside BioSynapStudio’s solver, but full credibility requires comparison against:
- Established simulators (e.g., NEURON, Brian2, NEST) under matched parameter sets
- Experimental electrophysiology from real neurons exhibiting similar bursting, rebound, or hyperexcitable regimes
- Alternative numerical methods (e.g., finite-volume, hybrid event-driven solvers, implicit methods)
We are specifically looking to collaborate with:
- Computational neuroscience labs interested in solver benchmarking
- Stem-cell / organoid teams who routinely work with hyperexcitable or rebound-capable neurons
- Electrophysiology groups generating HH-comparable spike trains
- Neuromorphic researchers evaluating biophysical fidelity vs. hardware constraints
- Modelling groups exploring channelopathies, intrinsic bursting, or excitability disorders
Interested groups can access the membrane traces, gating variables, current decompositions and phase-plane data shown here, and — where appropriate — run the same parameter sets inside their own modelling stacks to assess reproducibility.
By widening the circle of replication, we aim to formalise these early findings into a community-validated benchmark of HH-class behaviour under controlled biophysical perturbations.
📊 Benchmarking Against Established Frameworks
Qognetix will also benchmark these runs against results from established simulators such as NEURON, Brian2, and NEST, using identical parameter sets and solver configurations wherever possible. The goal is not to outperform them numerically, but to show that BioSynapStudio reproduces the same core electrophysiological phenomena — with the added benefit of unified solver stability and cross-domain transparency.
A forthcoming technical brief, “Emergent Pathophysiological Dynamics in a Biophysically Faithful Synthetic Neuron Engine,” will document these comparisons in detail, including membrane potential overlays, current decompositions, and parameter tables for reproducibility.
🌐 Sharing & Collaboration
To encourage community engagement, Qognetix plans to release an interactive demonstration through the BioSynapStudio Lab, where researchers can adjust leak potential and potassium conductance in real time and observe the transitions between quiet, spiking, and bursting states. This will be accompanied by downloadable benchmark datasets, allowing others to replicate or extend the findings independently.
🧭 The Broader Trajectory
These studies form part of a larger roadmap aimed at scaling from single-cell fidelity to network-level behaviour — exploring how collections of accurate synthetic neurons might exhibit collective phenomena such as synchrony, adaptation, or pathological spread. Understanding instability at the cellular level is a prerequisite for understanding intelligence at the systemic level.
By validating these effects with scientific rigour, Qognetix is laying the groundwork for a new class of synthetic systems — ones that think, adapt, and sometimes even fail, for the same reasons nature does.
Closing Line and Summary
Scientific progress often begins at the moment when a system behaves in a way we didn’t predict. For BioSynapStudio, that moment came when a mathematically perfect neuron began to show the same imperfections as a biological one — bursting, rebounding, and doublet spiking, all born from shifts in ionic balance.
These weren’t programming tricks or numerical artefacts. They were the natural consequences of biophysics expressed through a solver built to preserve, rather than suppress, instability. By allowing those instabilities to emerge, the model begins to reflect something fundamental about life itself: that complexity and vulnerability are two sides of the same coin.
When a synthetic neuron can fail for the same reasons a biological one does, it’s not just accurate — it’s alive in the only way a computation can be. It’s responding to the same physics that govern reality.
BioSynapStudio doesn’t simulate life; it rehearses it — faithfully, repeatably, and with just enough imperfection to remind us what makes biology remarkable.
