Reproducibility, built into the work — not bolted on.
Reproducibility takes four things: the same data, a record of how it was made, the ability to re-run to the same answer, and independent verifiability. In Radiens, all four are backed by a metadata and provenance chain that captures itself as you work — so reproducibility is a property of the workflow, not a chore you have to remember.
What reproducibility actually takes.
Four requirements — each backed by metadata and provenance, made first-class in Radiens rather than left to a lab notebook.
The same data
You can return to the exact recording behind a result — intact, identified, and still yours.
How it was made
The full transform history travels with the data — what was done, in what order, with which settings.
Re-run to the same answer
Pinned methods and a settable RNG let any result re-derive itself, exactly.
Independently verifiable
Open formats let someone else check the result without rebuilding your setup.
Today, reproducibility is a chore you have to remember.
Reproducibility usually rests on a person: someone has to write down what the experiment was, what each session and transform did, and keep that description stapled to the primary data. It fails in the ordinary human ways — the note that never got written (omission) and the value keyed in wrong, or drifted from what actually happened (transcription) — and it fails quietly at scale, across the hundreds of sessions a study accumulates. Radiens does not claim to make science reproducible on its own. What it does is make the reproducibility layer — the metadata and provenance chain — a first-class feature rather than an afterthought: captured automatically as the data moves through the pipeline, and linked to the data itself.
The record captures itself.
Every dataset, every transform — logged and linked to the primary data, automatically.Radiens auto-captures everything it can — the dataset’s identity, its signal-quality metrics, and every operation applied to it — and links that record to the primary data as it moves through the pipeline. The metadata a machine can know is never forgotten or mistyped; the descriptive metadata only you know is entered once, in structure, and bound to the data.
Dataset ID and provenance list
Every dataset carries a unique ID and a record of its transforms — the reproducibility foundation. Free.
Every transform, versioned
Each spike sort and DSP step — filtering, re-referencing, resampling — carries a unique ID with settable RNG, so any result re-derives itself exactly. Free.
Your metadata, entered once and bound
Attach session and study grouping and tags that travel with the primary data — described once, never re-keyed — and enrich for FAIR sharing (Pro).
You can always return to the same data.
Local-first, in your own storage, in open formats — nothing held hostage.Reproducibility requires the original data to still exist, unaltered, and reachable. Radiens works on your data where it lives, never takes custody, and reads it natively — so the exact recording behind a result is always there to return to, and the primary data is enriched, never replaced.
Local-first, your own storage
Work on data in place, or run compute against your own storage — GIN, DANDI, institutional S3 — with nothing uploaded to NeuroNexus. Free.
No vendor custody
Raw data never transits to NeuroNexus infrastructure; an IRB evaluates Radiens as a computational service, not a data custodian. Free.
Read natively, no conversion
Recordings from most acquisition systems, and common EEG/ECoG formats, open in place — the primary recording is enriched with metadata, never overwritten. Free.
Verifiable by someone who isn’t you.
Open standards out; the provenance carried further, on the roadmap.A result no one else can check is not reproducible. Radiens exports to open standards so anyone can verify a result without rebuilding your environment, and the provenance chain is headed toward a durable, verifiable record.
Export to open standards
Export to NWB, the open standard — portable and archive-ready, at every tier. Downgrading or canceling a subscription never locks you out of your data. Free.
FAIR-ready sharing
Enrich datasets with the metadata that makes them findable and reusable in public repositories (Pro).
Data certification
A durable record of a dataset’s origin and how it was produced — carrying the provenance chain further, for verification beyond your own lab.
The layer behind every result.
A clean acquisition, an analysis that holds up, a script or agent you can audit — each inherits the same chain. Reproducibility is not a page in Radiens; it is the substrate the whole workflow runs on.
Frequently asked questions
What makes a result reproducible in Radiens?
Reproducibility takes four things — the same data, a record of how it was made, the ability to re-run to the same answer, and independent verifiability — and Radiens backs each with metadata and provenance: every dataset carries a unique ID and versioned transform history, every operation is re-derivable with a settable RNG, the data stays in your own storage, and results export to the open NWB standard for anyone to check.
Does Radiens capture metadata and provenance automatically?
Yes. Radiens auto-captures the metadata it can — dataset identity, signal-quality metrics, and every signal transform such as filtering, re-referencing, and sorting — and links it to the primary data as the data moves through the pipeline, so the machine-knowable record cannot be omitted or mistyped. The descriptive metadata only you know is entered once, in structure, and travels bound to the data.
How is provenance different from reproducibility?
Provenance — the record of where a dataset came from and how it was transformed — is one component of reproducibility, not the whole of it. Reproducibility also requires that the original data still exists and is reachable, that a result can be re-run to the same answer, and that someone else can verify it; Radiens supports all of these, with the metadata and provenance chain as the connective tissue.
Is my data locked into Radiens?
No. Radiens reads your files natively with no conversion, works against your own storage with nothing uploaded to NeuroNexus, and exports to the open NWB standard at every tier — including free Standard. Downgrading or canceling never locks you out of your data.
What is data certification in Radiens?
Data certification is a roadmap capability that carries the provenance chain further — a durable record of a dataset’s origin and how it was produced, for verification beyond your own lab. Radiens already gives every dataset a unique ID and a versioned provenance list today; certification extends that foundation.
Reproducible by default. Your data stays home.
The provenance chain, dataset IDs, reproducible operations, and open export are there from the free tier — captured as you work, on your own storage. Reproducibility becomes a property of the workflow, not a chore.