What is TVB-O

Brain-network models are hard to reproduce and compare. Equations, parameters, networks, and numerical settings are reported inconsistently, and shared code drifts from the paper it came with. TVB-O fixes this by making the whole experiment one object.

You describe a simulation experiment once, as a typed specification grounded in an ontology. From that single source, TVB-O does two things: it compiles runnable code and a methods report, and it grounds every entity in a knowledge graph.

One specification

A specification captures everything needed to reproduce a run: the local dynamics, the network and its coupling, the integrator, any stimuli, the observation pipeline, and the provenance. Each part is written inline, loaded from the database, or referenced by a semantic pointer into the ontology.

Two payoffs

Reproducible. One compiler emits equivalent code across independent backends, plus a publication-ready methods report. A shared study re-runs without drift between what the paper says, what the code does, and what the docs claim.

Comparable. A four-domain ontology grounds each entity in physical units, biological process, brain anatomy, and clinical disorder. Two studies that use the same dynamics under different symbols become comparable by what their parameters mean, not by what they are called.

See it in action

Find and compare models in the Knowledge Graph:

Assemble and export a runnable experiment in the browser:

Run the same experiment in Python:

The pieces

Where to go next

• Getting Started — browse, build, run, and save.
• Knowledge Graph — search models, networks, couplings, and studies.
• Experiment Builder — assemble and run an experiment.
• Python package — run experiments in Python. Full docs at virtual-twin.github.io/tvbo.