Neuron Lab platform docs

Neuron Lab is a research platform for reproducible neurobioinformatics and computational neuroscience. It gives labs a single place to store datasets, launch analyses from a large module catalog, run foundation-model embeddings, and share results — with strict tenant isolation and a full audit trail.

Core building blocks:

• Projects — scoped containers for a research question.
• Datasets — uploaded or imported files (scRNA, GWAS, connectome, ephys, EEG, fMRI, other).
• Modules — analysis units (classical algorithms or foundation models).
• Analyses — single runs of a module against a dataset with saved parameters, live progress, and results.
• Batches — one module across many datasets/channels.
• Workspaces — shared teams with RBAC and webhooks.

1. Create an account at /auth (email + password).
2. Complete your profile (full name, institution, focus area, disease interests).
3. Go to Projects and click New project. Pick a domain — neurobio, compneuro, or mixed.
4. Open the project, switch to the Datasets tab, and upload a file or import from the public catalog (Allen, GTEx, HCP-style references).
5. Switch to the Analyses tab, pick a module, set parameters, and click Launch. Watch live progress on the analysis detail page.
6. When it completes, view results, export a PDF/CSV, or create a public share link.

Sign-in and sign-up live at /auth. Email + password is the only method enabled today; SSO (Google, Microsoft, SAML/SCIM) is available on Scale and Institution plans on request. On sign-up we capture your name, university, country, academic level, focus area (compneuro/neurobio/both), and disease interests. All authenticated pages sit behind the /_authenticated gate and redirect unauthenticated visitors back to /auth.

A project is the top-level scope for a research question. It holds datasets, analyses, and batches. Each project has a domain:

• neurobio — molecular / genomics workflows (scRNA, GWAS, sequence, multi-omics).
• compneuro — dynamical/network workflows (spiking models, spectral, connectome, plasticity).
• mixed — both domains available in the module picker.

Project counts are enforced by your plan (see §18). The Projects list shows a live usage banner and disables New project when you reach the cap.

Datasets are files scoped to a project. Supported kinds: scrna, gwas, connectome, ephys, eeg, fmri, other. On upload the server parses headers/metadata (format, sample rate, channels, duration) and stores a preview.

Two ways to add data:

1. Upload — drag-drop or file picker. EDF, CSV, TSV, and NIfTI are parsed automatically; other formats are stored as-is.
2. Import from public catalog — one-click references such as Allen Brain Atlas, GTEx, and HCP-style connectome samples.

Parse status is shown per dataset (pending → running → ready/failed). Failed parses expose the exact parser error so you can fix the header or file encoding.

Launching a run

From a project, open Analyses → New analysis. Pick a module — the list is filtered by the project's domain and the dataset's kind. Fill in the module's typed parameter form and click Launch.

Live status

The analysis detail page streams status/progress events as the run executes: queued → running → succeeded (or failed/cancelled). Failed and cancelled runs have a one-click Retry button that re-uses the original parameters.

Results

Results are rendered by module-specific viewers — spectrograms, PCA/UMAP embeddings, Manhattan plots, connectome heatmaps, spike rasters, etc. — plus:

• Plain-language interpretation summarising what the numbers mean.
• QC panel flagging inputs that fall outside your configured thresholds.
• Timeline of every status event and parameter change.
• Export as PDF/CSV, or open the clinical-style report (§11).
• Share a read-only public link (§13).

The full module catalog lives at /modules. Modules are grouped into two domains:

Neurobioinformatics

• Single-cell & bulk RNA-seq — QC, normalization, HVGs, PCA, clustering, DEGs, marker discovery.
• GWAS & population genetics — association testing, LD, PCA, kinship, heritability.
• Sequence & variant analysis — alignment stats, variant calling summaries, annotation.
• Multi-omics integration, pathway enrichment, network inference.
• Foundation models for genomics and single-cell (embed, cell-type inference, perturbation).

Computational neuroscience

• Neuron models — Hodgkin-Huxley, LIF, AdEx, Izhikevich, Morris-Lecar, Hindmarsh-Rose.
• Spike-train statistics; network dynamics — Wilson-Cowan, Kuramoto.
• Plasticity & memory — STDP, Hopfield.
• Signal analyses — spectral, ERP, coherence, connectome graph metrics.

Every module declares its accepted dataset kinds and a typed parameter schema, so the launcher can only offer valid combinations and can render a correct form for you.

Three foundation models share the same launcher, storage, and viewer pipeline as classical modules:

• Cell Atlas FM (cell_fm) — scRNA embedding with zero-shot annotation and clustering.
• Neural Signal FM (signal_fm) — EEG/ephys embedding with dynamical-state classification.
• Connectome FM (graph_fm) — spectral node embedding and community detection.

Launch from /foundation-models (marketing) or the in-app Foundation section. Backbones are swappable per model. Runs appear alongside classical analyses; the FM-only jobs view filters to running / succeeded / failed FM runs.

Batches run one module across many datasets or channels in bulk. Create a batch, pick the module + parameters, select target datasets/channels, and queue it. The batch detail view shows per-item status and lets you cancel in-flight items or export the batch as a CSV report.

Batch analyses are available on Lab and above.

Every analysis is compared against a benchmark corpus derived from published results (MNE, EEGLAB, FieldTrip, YASA, Elephant, and others) on canonical public-dataset fixtures. Each run gets a Research confidence score that:

• Matches your run's metrics to benchmark rows for the same module/dataset kind.
• Reports deviation % vs. the reference value.
• Flags likely causes when deviation exceeds your thresholds (parameter drift, input QC, sample size).
• Lists unmatched metrics that need review.

The public /benchmarks page runs the same engine in-browser against a small demo fixture so you can validate the numbers yourself.

Any analysis can be rendered as a structured clinical-style report with subject, ordering, findings, panels, and figures. From the report page you can export:

• PDF — print-styled, ready for institutional records.
• JSON — the raw report object.
• FHIR bundle — for EHR integration.
• HL7v2 ORU — for lab-message pipelines.
• CSV — for spreadsheet analysis.

The in-app assistant lives at /chat. Create threads scoped to a project or dataset, ask questions about your data, get parameter-picking help, or ask for explanations of results. Threads are stored per user and can be deleted at any time.

From an analysis detail page, click Share to mint a read-only token. The recipient opens /shared/<token> and sees the module name, status, parameters, and every result panel — no login required.

Options:

• Expiry — auto-revoke after N days.
• Revoke — manually kill any token at any time.

Expired or revoked tokens render a friendly error page.

Workspaces are shared teams. Invite members by email and assign a role:

• Owner — full control, billing, delete workspace.
• Admin — manage members, projects, webhooks.
• Member — create and run analyses.
• Viewer — read-only access to results.

Projects can be personal or scoped to a workspace. Workspace resources appear for every member subject to their role and RLS.

Configure signed outbound webhooks per workspace at /workspace. Each webhook receives HMAC-signed events:

• analysis.succeeded — includes analysis id, module key, project id, and results summary.
• analysis.failed — includes error message and last event.

Verify signatures with the shared secret shown once at creation time. Requests are retried on 5xx responses with exponential backoff.

RBAC is stored in a dedicated user_roles table (admin / moderator / user) and checked server-side via the has_role function. Admins get an in-app console at /admin with:

• Neuron Lab-wide stats (users, projects, datasets, analyses).
• User list with role grant/revoke/delete.
• Plan management.
• Tables to manage projects, datasets, and analyses across all tenants.
• Full audit log at /audit.

Every tenant can override input-quality thresholds and confidence-score bands at /settings/qc, plus per-(module, metric) range overrides. Runs use your overrides when flagging QC panels and computing research confidence.

Full pricing at /pricing. Summary:

• Explore — free. 1 project, 1 analysis/month, all modules, public datasets, community support.
• Starter — $29/user/mo. 5 projects, 100 analyses/mo, PDF/CSV export, share links, email support.
• Lab — $99/user/mo. 25 projects, 1,000 analyses/mo, batches, webhooks & REST API, workspaces + RBAC, audit logs.
• Scale — $299/user/mo. 100 projects, 10,000 analyses/mo, benchmark ingest, SSO, SOC 2, Slack support.
• Institution — custom. Unlimited usage, SAML/SCIM, dedicated compute, custom modules, SLA/DPA, named CSM.

Plan usage (projects + monthly analyses) is shown on the dashboard and on the Projects list.

Every user-scoped table is protected by row-level security keyed on created_by, owner_id, or user_id. Roles are stored separately from profiles to prevent privilege escalation. Key tables:

• projects, datasets, analyses, analysis_results, analysis_events.
• analysis_shares — public share tokens with expiry/revocation.
• analysis_batches, analysis_batch_items.
• workspaces, workspace_members, workspace_webhooks.
• chat_threads, chat_messages, notifications.
• profiles, subscriptions, user_roles.
• benchmark_runs, audit_log.

Server-side helpers include current_user_plan, plan_limits, has_role, is_workspace_member, and get_shared_analysis.

The /diseases page maps 15 disease classifications (neurodegenerative, cerebrovascular, epileptic, movement, psychiatric, and more) to the specific modules most useful for each — e.g. epilepsy → Wilson-Cowan, Hodgkin-Huxley, and spectral analyses. Use it to jump straight from a research question into the right launcher.

Reach us via the /contact page or the email channels listed there (hello / support / sales). Community support is included on all plans; Lab and above get email support, Scale adds Slack, and Institution includes a named CSM with an SLA.

Common questions are answered on the /faq page.