We earn trust where mistakes are expensive.

A platform this broad has to prove itself in the room with the highest stakes. So we picked that room first.

We started with cardiology.

On purpose. A wrong number there shows up as a body count, and everyone in the field knows the numbers by heart. If we couldn't be honest in the hardest place, no one would trust the rest.

The same engine runs every domain above.

Cardiology is just where the receipts are longest. More published trials, more endpoints, more ways to check whether we are right or just sounding right.

Nothing was fitted to the answer.

The whole thing builds from physics: molecular structure, to binding kinetics, to organ-level electrophysiology. Each layer follows from the one before it, not from the result we wanted.

Then the layers connect.

Change a molecule and the signal moves through receptors, cells, tissue, organs, and the whole person. The mechanism is the answer, not just the final number.

We checked it against 72+ real trials.

Cardiology, ARDS, stroke, sepsis, oncology. The scorecard is public, the assumptions are open, and the gaps are named honestly.

Have a look for yourself.

Open the live demos, browse the trial evidence, or just read how the pipeline works. Then come talk to us about what you want to model.

See the evidence How it works

Scroll to build

01 What you can ask it

One simulation engine, many practical questions.

oNeura is designed to connect a molecule to the body in a traceable way. Instead of treating each disease as a separate model, the same pipeline can estimate properties, target effects, organ response, and downstream outcomes. A few examples:

Drug discovery

From a structure to an effect

Enter a SMILES string. The model estimates the molecule's properties, how it may bind to targets, and what downstream effect that could have. It is meant to explain the mechanism, not just return a score.

Cardiac safety

Catch heart-rhythm risk earlier

Estimate hERG, Nav1.5 and Cav1.2 effects, then run those effects through a heart-cell model. The result is a clearer view of pro-arrhythmic risk from the molecule's structure.

Disease modeling

Watch a disease unfold

Conditions like ARDS, sepsis, stroke, kidney injury and liver failure can be represented as interacting body systems. That makes it easier to ask what changes when a drug, ventilator setting or disease process changes.

Neuroscience

Understand neural circuits

Simulate neural activity from receptor signaling and published circuit parameters. This helps show how molecular and cellular changes can lead to measurable brain activity.

Oncology

Trace tumor growth from signals

Model cell-cycle control, EGFR and AR signaling, antibody-drug conjugates and invasion patterns as connected mechanisms rather than isolated endpoints.

Longevity and metabolism

Study energy, stress and aging pathways

The same engine can represent mitochondrial energy use, oxidative stress, ferroptosis, and metabolic or endocrine pathways that change over time.

Immunology

Connect immune signals to whole-body response

Model adaptive immunity, complement activation, cytokine storms and checkpoint signaling as part of the same body loop.

Precision medicine

Explore one patient's physiology

A patient-specific model can test ventilator settings, drug regimens and projected outcomes for that body in particular, while keeping the assumptions visible.

02 Where we proved it first

We tested it against known clinical outcomes.

The clearest place to start was cardiology because heart outcomes are well measured and mistakes are consequential. We used the same modeling pipeline for other domains, but cardiology gives us the most direct comparisons against published trial results. The examples below show how a molecule can be taken from structure through binding, cell behavior, organ response and body-level readout.

Cardiology, first proof: the PLATO trial

We used ticagrelor's molecular structure to predict the trial's mortality outcome.

The PLATO trial enrolled 18,624 patients across 43 countries at a cost exceeding $500M. Our simulation produced the same answer from a SMILES string. No patient data, no outcome fitting, no lookup.

Clinical Trial Result

9.8%

oNeura Prediction

within 2pp

Input: ticagrelor SMILES string only. Pipeline: molecular properties → P2Y12 binding estimates → PK/PD → cardiac electrophysiology (O'Hara-Rudy) → hemodynamic outcome. The assumptions are visible at each step.

72+

trial arms compared across cardiology, pulmonary, neurology and oncology. The goal is to show where the model matches known results and where more validation is needed.

Cardiology 17/18

PLATOwithin 2pp

RE-LYwithin 2pp

ARISTOTLEwithin 2pp

ENGAGE-AFwithin 2pp

ROCKET-AFwithin 2pp

AVERROESwithin 2pp

+ 11 more+/- 2pp

ARDS / Ventilation 10/10

ARMAwithin 5pp

ACURASYSwithin 5pp

PROSEVAwithin 5pp

FACTTwithin 5pp

ROSEwithin 5pp

+ 5 more+/- 5pp

Neurology 10/10

Epilepsy modelswithin 5pp

Parkinsonian MPTPwithin 5pp

Stroke thrombolysiswithin 5pp

Alzheimer's AChEIwithin 5pp

+ 6 more+/- 5pp

Oncology 25/35

Chemotherapy PK/PDvalidated

Tumor growth dynamicsvalidated

Combination therapyin progress

Resistance modelingin progress

+ 22 validated25/35 total

pp = percentage points. Trial comparisons use the oNeura multi-organ simulation pipeline. Drug properties are derived from molecular structure. Outcomes are generated by connecting molecular, cellular, organ and whole-body models. Cardiac electrophysiology uses the O'Hara-Rudy ventricular myocyte model and a CiPA-aligned ion channel panel (hERG, Nav1.5, Cav1.2, KCNQ1+minK, Kir2.1). These results are research outputs, not clinical advice.

03 How people use it

From a question to an answer you can discuss.

The platform is broad, but most teams come to us with concrete decisions. Here are the jobs we are building for first.

Screen safety before synthesis

Run a candidate's cardiac liability from its SMILES string. The model estimates hERG, Nav1.5 and Cav1.2 effects and returns a CiPA-style readout so teams can rank molecules before bench work.

Cardiac and CiPA stack

Ask a mechanism question clearly

When the question is about how a pathway behaves, simulate the pathway and show the assumptions. Each output can be traced back through the model rather than hidden in a black-box score.

Atom to organ pipeline

Build an evidence package

Verification and validation suites produce the documentation a submission or internal review needs. Runs are deterministic and hash verified, so results can be reproduced later.

V and V 40 suites, deterministic runs

Explore a patient-specific case

A REST and WebSocket API steps a single ICU patient forward. Change the ventilator, add a drug, read the vitals, SOFA score and projected outcome for that case.

Digital twin API

Design peptide or antibody therapeutics

Go from a sequence to estimated properties, simulated trial behavior, drug-to-antibody ratio sweeps and de-immunization suggestions using the same modeling engine.

Peptide and therapeutic APIs

Follow disease from molecule to organ

Perturb one molecule and watch the effect move through binding, cell state, tissue, organ function and whole-body response. The mechanism is part of the answer, not just the final number.

Closed-loop multi-organ simulation

04 How it works

From a molecule to a body-level readout.

We start with the molecule, estimate how it behaves, then pass those results into connected organ models. The point is to keep the chain of reasoning visible: molecule, target, cell, organ, body.

01 / PARSE

Molecular structure

Enter any drug as a SMILES string. We parse the molecular graph and compute basic properties such as molecular weight, LogP (Wildman-Crippen), polar surface area, and hydrogen bond donors/acceptors.

SMILES LogP PSA Lipinski

02 / BIND

Binding estimates

Drug-target binding rates are estimated from molecular properties and physical encounter models. Kd, kon and koff are produced by the model so the path from structure to effect stays visible.

Smoluchowski Eyring TST Kd estimate

03 / SIMULATE

Connected body model

Binding drives receptor occupancy, occupancy drives organ function, and organ function drives whole-body regulation. Heart, brain, lung, liver and kidney are coupled so a change in one place can be followed through the rest of the system.

O'Hara-Rudy CiPA panel ECG output QTc

Atom Basic atomic properties and bond behavior used as the starting point for the model.

Molecular Molecular shape, motion and descriptors that describe how a compound behaves in solution.

Compound SMILES parsing, molecular properties, PK/PD estimates and pharmacogenomic context.

Cellular Cell-state models, signaling lattices and immune-cell behavior.

Tissue Tissue mechanics, extracellular matrix, spatial cell populations and tumor microenvironment.

Organ Heart, brain, lung, liver, kidney and immune modules wired into HblSim::step().

Organism Closed-loop heart-brain-lung regulation with baroreflex, chemoreflex and homeostatic feedback.

05 Multi-species validation

Signals come from biology, not shortcuts.

We model how receptors and molecular cascades drive cell activity, instead of tuning the model to match a desired output. Each parameter is tied to a source or derivation so the reasoning can be reviewed.

C. elegans

302 neurons · 2.1 Hz mean firing

GPCR-cGMP-TAX-2/4 cascade. 2 μM diacetyl at AWA produces sparse firing matching published rates. Complete connectome from White 1986, validated against Cook 2019 Nature 571:63.

Drosophila melanogaster

25,000 neurons · Laptop25K class-balanced

Or/Orco ionotropic cascade (Sato 2008, Wicher 2008) with DoOR v2.0 receptor Kd values. PN firing 60–150 Hz, KC firing 0.1–20 Hz matching Wilson 2004 and Turner 2008. Spectral peaks at 38 Hz (gamma) and 162 Hz from connectome structure alone.

Mouse V1 cortical column

10,000 neurons · MICrONS EM connectome

GPCR–Gαolf–cAMP–CNG cascade (Bhandawat 2005, Kaupp 2002) drives thalamocortical afferents. L4 stellate drive from Bruno & Sakmann 2006. Firing rates validated against Lefort 2009 and Schneider-Mizell 2024 Cell 187.

V&V40 Credibility

≥ 0.90

ASME V&V40 standard credibility score across four organ probes (cardiac, neural, pulmonary, hepatic). Intended as validation evidence for review, not a substitute for clinical validation.

Test coverage

2.27M lines of Rust across 53 crates. Every layer tested from unit through integration. CI-regression-gated pull requests prevent biology violations from merging.

06 Live evidence

Open the live demos

These demos run simplified versions of the pipeline in your browser. They are meant to make the logic visible, not replace clinical judgment.

Simulation

ICU Sepsis Simulator

Follow a 72-hour sepsis trajectory and change fluids, vasopressors, antibiotics or CRRT. The simulator shows how common ICU interventions can move MAP, lactate, SOFA score and survival probability.

Launch simulator →

Pharmacology

Drug-Drug Interaction Engine

Choose 2-3 drugs from 12 real medications. The demo shows CYP enzyme competition, protein binding displacement and additive pharmacodynamic effects, including the Warfarin + Fluconazole CYP2C9 conflict.

Launch engine →

Evidence

72-Trial Evidence Explorer

Browse 72 clinical trial-arm comparisons across cardiology, ARDS, stroke, sepsis and oncology. Filter by domain, trial phase and fidelity score to see where the model matches known results.

Explore evidence →

Pipeline

SMILES Drug Calculator

Enter a drug SMILES string and get molecular weight, LogP, PSA and Lipinski drug-likeness. This is the first step in the larger SMILES-to-clinical-outcome pipeline.

Calculate →

07 How to work with us

Ways we can work together

Tell us what you are trying to learn, decide or validate. We can start with a focused pilot and grow from there.

Research

Academic & Government

Source-available simulation engine for computational biology research. Useful for NIH SBIR/STTR proposals, university collaborations and government-funded research programs.

Full platform access for research use
Co-authorship on publications
Grant application support

Pharma

Drug Screening & Safety

Screen drug candidates for cardiac toxicity, drug-drug interactions and clinical outcomes before running a trial. The output can support internal review and regulatory discussions.

SMILES-to-ECG cardiac safety pipeline
Multi-drug interaction screening
Regulatory evidence packages (V&V40)

Defense

CBRN & Medical Readiness

Simulate whole-body response to chemical agents, radiation exposure and multi-trauma. Generate training scenarios and compare treatment approaches for defense medical research.

Multi-organ failure simulation
Treatment protocol optimization
ARO/ONR/DARPA research support

08 Partnership

Tell us what you are trying to solve how a molecule may affect a body

oNeura is developed by Blackweb AI (Florida, UEI Y7QQLFJKZ1H4). We are open to investment under our Delaware C corporation, Blackweb AI, Inc., and we are looking for pharma pilots, federal grant partners, accelerator programs and academic collaborators.

Robert Price

Founder & Creator

Robert created oNeura and leads the platform architecture. He is a University of Florida graduate with a B.S. in Computer Science and focuses on the multi-organ simulation engine, the SMILES-to-outcome drug pipeline and CiPA-aligned cardiac safety methods. LinkedIn → · GitHub →

Muntaser Syed

Lead Researcher · PI, Neural Architecture Search

Muntaser is an AI/ML researcher and PhD candidate at Florida Institute of Technology. He leads neural architecture search applications for oNeura, with a focus on energy-efficient machine learning on edge devices and trustworthy AI. Dissertation: Artificial Intelligence of Things: Effective Machine Learning on Edge Devices. IEEE and ACM member. LinkedIn → · GitHub →

Get in touch

bobby@blackweb.ai

inquiries@blackweb.ai

blackweb.ai/contact

ARPA-H IGoR proposals
NIH SBIR / STTR applications
Academic consortium partnerships
Defense research (ARO / ONR / DARPA)
Pharma licensing and co-development
Investment (Blackweb AI, Inc., Delaware C corp)

Request a pilot study

We'll run your drug candidate through the SMILES-to-outcome pipeline and share a cardiac safety assessment with the assumptions shown.

Join the source waitlist

2.27M lines of Rust, 19K+ tests, 53 crates. Source access is in closed beta for design partners. Reach out if you want to evaluate it with a real use case.

Browse the evidence

72 trial arms across 7 domains, filterable by domain, phase and fidelity score.

72 Trials →

Invest in oNeura

We're open to investment under Blackweb AI, Inc., a Delaware C corporation. Reach out for the data room and pitch materials.