Neuromembrane Beta 2.0

Neuromembrane 1.0 used an ambitious layered canvas system to animate a library of carefully illustrated biological components. Version 2.0 keeps that commitment to making invisible physiology visible, but rebuilds the visual layer as a real-time environment.

Channels, pumps, receptors, ions, vesicles, electrodes, and synaptic structures now exist as interactive parts of a coordinated scene. Their state follows the simulation timeline: gates move, ions flow, ligands travel, magnesium block changes, and membrane charge responds as the underlying data changes.

Every simulation also has a complementary 2D treatment. These are not screenshots of the 3D scene; they are schematics designed to clarify topology, current direction, receptor state, and equivalent-circuit relationships when spatial depth would get in the way.

The new simulator is organized around a persistent experimental workspace. Parameters sit beside the visualization, graphs remain available beside both, and the playback clock gives every surface a shared sense of time.

Moving the scrubber can update the scene and the graph cursor together. Graph panels can be collapsed or reordered, the settings are grouped into resettable sections, and protocol editors make multi-step current and voltage experiments legible before they run.

On smaller screens, that workspace becomes a focused control island with dedicated Parameters, Graphs, Summary, and Learning views. It is the same simulator rather than a reduced mobile substitute.

Beta 2.0 extends the action-potential experiment with Channelrhodopsin-2. Students can add an optogenetic conductance, define light protocols, and study how light-driven current interacts with the familiar Hodgkin–Huxley sodium and potassium dynamics.

The implementation uses a four-state ChR2 model with voltage-dependent rectification. Its current, state transitions, channel appearance, and ion movement are carried through the solver, graphs, 3D scene, 2D schematic, tooltips, and tutorial-authoring vocabulary. This is a work-in-progress that is undergoing testing and validation.

The learning material in 2.0 is no longer confined to a help panel. A tutorial can point to an interface control, highlight a structure in the scene, adjust initial conditions, control playback, and wait until the learner has completed a meaningful action.

Tutorial steps can ask questions with answer-specific feedback or require conditions such as reaching a time point, changing a parameter, reversing ion flow, comparing a Nernst potential, or editing a protocol step. The lesson advances with the experiment instead of merely describing it from the side.

A built-in visual authoring environment makes the same system available to instructors and other creators. Tutorials can be drafted, previewed against the live simulator, published, linked directly, and completed with a signed participation receipt.

Signed-in learners can ask Neuromembrane to analyze the result of an experiment. Rather than returning a detached paragraph, the summary is organized as a sequence of time-bounded observations.

Selecting an observation seeks the simulation to the relevant moment. During playback, the active observation follows the timeline and can highlight the structures involved, connecting an explanation to the graph and biological scene it describes.

These summaries are generated assistance, not an answer key or a substitute for instruction. The interface marks them as AI-generated and makes it clear when parameter changes have made an existing analysis stale.

The original Neuromembrane was built on Python 2.7, Django 1.9, server templates, jQuery, Highcharts, and a large raster-asset pipeline. That architecture carried the simulator for years, but made it increasingly difficult to extend, test, secure, and deploy.

Version 2.0 separates a typed Next.js and React interface from a FastAPI simulation service. Scientific inputs are validated at the API boundary, expensive solver work is bounded, database changes are versioned, and optional Redis caching can share results safely across processes.

The equations remain the heart of the product. The rebuild gives us a more reliable way to protect them: legacy-parity fixtures compare representative outputs with 1.0, focused test suites cover each simulation family, and the frontend uses explicit result contracts instead of relying on loosely shaped response objects.