A multiscale simulation platform linking molecules, neurons, and behavior
oNeura combines browser-live biophysics, replayed native circuit experiments, and a growing molecule-to-circuit pharmacology stack. This page distinguishes what is live in the browser, what is replayed from native GPU runs, and what is under active integration for NeuPharma.
Status guide: live in browser = interactive simulation; replay = measured native run visualized on the web; integration = implemented locally but not yet fully exposed on this site.
oNeura is a physics-native neural simulation platform. Publicly visible slices already span a browser-live Hodgkin-Huxley neuron, replayed native Pong experiments, and fly-brain-scale assays measured on the Rust + CUDA runtime. NeuPharma is the commercialization layer that extends this stack upward from explicit molecules into pharmacology-aware neural and behavioral workflows.
The important distinction is scope. Some surfaces on this site are live in your browser, some are faithful visualizations of native runs that are too large for a browser, and some molecule-to-circuit workflows are implemented locally and being prepared for public exposure. That separation is deliberate so claims stay tied to current evidence.
Single-neuron Hodgkin-Huxley dynamics, browser Pong, WebGPU fly-brain demo, and Y-maze assay interfaces that expose the same target classes and circuit motifs as the native stack.
25,250-neuron Pong results and 139K-neuron fly assays are shown with measured outputs from the native Rust + CUDA runtime when full-scale execution exceeds practical browser limits.
NeuPharma connects explicit molecules, docking, MM-GBSA calibration, membrane permeation, and occupancy-informed neural perturbation into one commercializable pharmacology workflow.
This browser panel now defaults to the mammalian cortical kinetics preset from oNeura’s native channel_kinetics_type system: Nav1.2-like sodium activation and Kv3.1-like potassium repolarization instead of the old universal squid baseline. The membrane potential still emerges from ion-channel physics, while the Rust pharmacology path initializes supported compounds from explicit molecules, calibrates affinity with docking and MM-GBSA, and updates that baseline with live ligand MD before target occupancy changes neural state.
Browser preset: mammalian cortical Nav1.2 / Kv3.1 kinetics. Fly demos on this site use the Drosophila para / Shaker preset.
Visual replay of recorded native experiment data. The 25,250-neuron brain is too large to run live in a browser, so what you see below is a faithful replay of actual outcomes across 80 rallies. The same brain architecture runs identically on a laptop, a datacenter GPU, or anything in between - oNeura is hardware-agnostic. Toggle below to see real results from each platform.
Measured March 10, 2026 from native comparison runs. Random baseline = 30%. All runs: seed 42, 80 rallies, dt=0.1ms, 12 frames/rally. 25,250 neurons, ~7M synapses.
A 5,050-neuron spiking brain sees Doom through a molecular retina, processes visual input through STDP-learning cortical circuits, and outputs motor commands to move, turn, and shoot. The brain is first shaped through 12 teacher-guided episodes that co-stimulate the correct motor populations, then evaluated in autonomous play without a handcrafted decoder policy. Treat this as an early behavioral benchmark rather than the primary proof point for the platform.
Teacher pretraining builds synaptic connections via STDP by stimulating correct motor populations alongside visual input. During autonomous play, the brain chooses actions solely from L5 cortical spike counts - whichever motor population fires most wins. No biases, no shaping, no heuristics.
These combat runs use the same target classes that matter in real circuits. In the native stack, supported compounds now enter through explicit molecule graphs, surrogate-pocket docking, MM-GBSA calibration, membrane permeation, and ligand-MD-perturbed occupancy before they modulate conductance or transmitter state. Caffeine increases excitability, diazepam increases inhibition, and amphetamine boosts catecholamine signaling.
Measured from native VizDoom combat runs on RTX A6000. 5,050 neurons, 308K synapses, defend_the_center scenario. Pharmacology data comes from the same native combat pipeline.
Real experiment results from a 139,000-neuron fly brain simulation with 10 million synapses, running on an NVIDIA RTX A6000 (46GB VRAM) with our native Rust + CUDA backend. Every neuron runs the full Hodgkin-Huxley biophysics with ion channels, receptor binding, calcium dynamics, and STDP plasticity - all computed in parallel on the GPU.
1,000 photoreceptor neurons stimulated at 50 µA/cm². Network firing rate increases from 0 Hz baseline to 0.40 Hz, then decays to 0.28 Hz during recovery.
Graded temperature ramp (18°C–34°C). Network shows initial excitation followed by sensory adaptation at higher temperatures.
Caffeine (Nav +30%) increases firing rate by 7.8%. GABA agonist and dopamine modulation operate through downstream molecular pathways.
Backend: Rust + CUDA 12.4 (cudarc). HH gating, synaptic transmission, STDP plasticity, and NT dynamics all computed on GPU.
Measured from the native Rust + CUDA experiment pipeline and GPU brain runtime.
The older site copy described oNeura as 25 molecular subsystems per neuron. The current oNeura repo now presents the flagship Drosophila dONN path as materially deeper: roughly 173 state variables per neuron flowing through an approximately 45-stage pipeline, with many more ion-channel, signaling, structural, organelle, and gene-regulation layers than this page originally summarized.
The current repo framing now calls out 18+ channel classes, including Nav, Kv, Kleak, Cav, NMDA, AMPA, GABA-A, nAChR, HCN, SK, BK, TRPC, T-Cav, IA, GIRK, K-ATP, KCNQ, and K2P, with species-specific kinetics for squid, Drosophila, and mammalian presets.
12+ signaling cascades, 4-compartment calcium with SOCE and buffers, multi-ion concentration tracking, neurotransmitter release/reuptake, organelle state, and much broader intracellular bookkeeping now appear in the repo-level dONN description.
3-factor STDP is still central, but the current repo story is broader than the old summary: receptor trafficking, synaptic tagging, metaplasticity, neuromodulatory eligibility traces, and downstream structural/molecular consequences all sit inside the same neuron-and-circuit path.
Supported native compounds now bridge explicit molecules, surrogate-pocket docking, MM-GBSA calibration, membrane permeation, ligand MD, and occupancy-driven neural effects. The browser panels expose the same target classes and resulting circuit phenotypes.
The current repo narrative now explicitly includes histone acetylation, DNA methylation, miRNA/promoter occupancy, c-Fos/Arc/BDNF expression, ribosome/mTOR translation gating, astrocytes, oligodendrocytes, microglia, actin/spine dynamics, microtubules, neurofilaments, and extracellular-matrix state.
The repo now publicly describes a seven-scale stack spanning quantum, atomistic, cellular, stochastic, organismal, ecological, and evolutionary levels, plus a large Rust terrarium architecture with 263K+ lines, 287 source files, and 2,100+ tests. The website should be read against that broader project reality.
The goal of the stack is an unbroken causal chain from molecular perturbation to circuit and behavioral effect. Public pages currently expose selected validated slices of that chain rather than claiming every layer is fully public at once.
oNeura is the scientific engine: explicit neural state, measured experiments, and multiscale biology infrastructure. NeuPharma is the applied layer for molecule-to-circuit pharmacology, drug-discovery studies, and partner-facing simulation workflows. The public site focuses on validated neural and organism slices while the molecule workbench is being productized.
The core asset is not one demo. It is the shared simulation substrate that lets molecular and pharmacological changes propagate into neural and behavioral assays without switching toolchains at every layer.
The public surfaces separate browser-live simulations, replayed native runs, and under-development integrations so experimental claims remain tied to methods, hardware, and measured outputs.
NeuPharma productizes the molecule workbench, target-binding path, and assay workflows around oNeura so neuropharmacology and drug-discovery studies can move from structure to circuit effect inside one system.