Forward scatter against side scatter for 10,152 events from acquisition SRC0003, rendered as an orange density map on black.

SSC-A ↑FSC-A →

SRC0003.fcs · BD LSRFortessa
10,152 events · sha256 2d07b61c…
FSC-A 99,615 · SSC-A 99,615

Source-linked wet-lab data

We run the experiments, and deliver them as data a model can be held to.

Flow cytometry, Western blot, CCK-8, LDH, RT-qPCR and ELISA produced in contract labs. Every value traces to its workbook cell. Every raw instrument file is hashed. Generated annotations never mix with what the lab recorded.

Request the free preview Explore the data

Rendered from the raw FCS file, not a stock image · log density · cursor reads instrument unitsPilot 0.1.1 · counts as of 4 Nov 2026

Annexin V against propidium iodide density map from the same acquisition.

330

recorded measurements, each pointing to a workbook cell

54FCS

flow acquisitions, 540,152 events, raw and as Parquet

233

source files with SHA-256 hashes in a registry

Free preview package324 KB

sample_measurements.jsonl30 records in the full preview, 12 shown on this pageJSONL

sample_10_flow_events.parquetten real events, SRC0003Parquet

sample_wb.pngone Western blot export, 1830 × 1464PNG

sample_task_requests.jsonlwith reference candidates and scorerJSONL

buyer_brief.htmlinventory, counting boundaries, statusHTML

Request it

Every value is a cell, not a claim.

Three recorded values per condition, read straight from the lab workbook and kept as recorded. No averaging, no cleaning, no inferred replicates. The chart is drawn from the twelve preview rows on this page.

See the rows

Recorded pairs per assay, each pair scaled to its own maximum · units as recorded · no statistics implied

Every file is hashed and typed.

The registry lists all 233 source files with SHA-256, byte size, role and rights status. Raw acquisitions and exports are delivered as they left the instrument, not re-encoded.

What is in the archive

Every annotation says where it came from.

Captions, task references and context summaries are a separate layer with their own origin and review fields. Nothing generated is written back into recorded values.

How review works

{ "measurement_id": "P2.F2C.si_NC_ctrl_t0.r1", "value": 0.205, "unit": "OD450", "source": { "source_id": "SRC0185", "sheet": "CCK8 si", "cell": "B3" }, "value_provenance_status": "verified_against_source_cell", "annotation_origin": "source_extraction_and_programmatic_curation", "human_annotation_review_status": "unreviewed", "quality_flags": ["biological_independence_unverified", "inferred_layout", …], "leakage_group_id": "FLNA-STUDY-001", "split": "development_only" }

One record from sample_measurements.jsonl, fields abbreviated

What you get

Three layers, delivered apart

The measurements

One row per recorded value: assay, endpoint, condition group, cell model, replicate index, unit, and the workbook, worksheet and cell it came from.

CCK-8, LDH, ELISA optical densities
RT-qPCR Ct paired with reference gene
Flow percentages and blot densitometry as recorded
38 calibration standard points

measurements.jsonl · .csv · .parquet
contexts.jsonl · schema.json · validate.py

The files

Original acquisitions and exports. Every file has a registry entry with hash, byte size, role and rights status.

54 FCS 3.0 acquisitions + one Parquet each
51 primary blot PNGs, 153 TIFF export variants
6 workbooks, 6 Prism projects, 6 figure decks
2 lab-record documents

SOURCE_REGISTRY.jsonl · 233 entries
sources/SRC0001 … SRC0233 · 1.2 GB

The tasks

Machine-generated evaluation items over the measurements: value extraction, descriptive arithmetic, evidence-boundary judgement. Expected answer and tolerance on each.

263 task candidates with executable scorer
Leakage group and split on every item
Expert annotation as a separate service

task_requests.jsonl · reference_candidates.jsonl
score.py · human_verified: false

sample_measurements.jsonl · 12 rows from the preview package, values as recordedre-read against raw workbooks 4 Nov 2026 · 12/12 match

measurement_id	assay	group	value	unit	source cell	provenance
P2.F2C.si_NC_ctrl_t0.r1	CCK-8	si-NC control · astrocyte	0.205	OD450	SRC0185 · CCK8 si!B3	verified
P2.F2C.si_NC_ctrl_t0.r2	CCK-8	si-NC control · astrocyte	0.226	OD450	SRC0185 · CCK8 si!C3	verified
P4.F4A.oe_NC.r1	ELISA · MIF	oe-NC · astrocyte	0.155	OD	SRC0233 · MIF ELISA!A20	verified
P4.F4A.oe_NC.r2	ELISA · MIF	oe-NC · astrocyte	0.102	OD	SRC0233 · MIF ELISA!A21	verified
P1.F1A.GSC_ctrl.r1	Flow · Annexin/PI	GSC control	2.26	%	SRC0054 · apo!A3	verified
P1.F1A.GSC_ctrl.r2	Flow · Annexin/PI	GSC control	2.39	%	SRC0054 · apo!A4	verified
P2.F2B.si_NC_ctrl.r1	LDH release	si-NC control · astrocyte	0.029	OD	SRC0185 · ldh si!A22	verified
P2.F2B.si_NC_ctrl.r2	LDH release	si-NC control · astrocyte	0.074	OD	SRC0185 · ldh si!A23	verified
P1.F1E.GSC_ctrl.r1	WB densitometry · SLC7A11/GD	GSC	226,557	intensity	SRC0054 · slc7a11!B2	verified
P1.F1E.GSC_ctrl.r2	WB densitometry · SLC7A11/GD	GSC	225,997	intensity	SRC0054 · slc7a11!B3	verified
P2.F2A.si_NC.r1	qPCR · FLNA/GD	si-NC · astrocyte	26.16	Ct	SRC0185 · si PCR!B3	verified
P2.F2A.si_FLNA.r1	qPCR · FLNA/GD	si-FLNA · astrocyte	27.55	Ct	SRC0185 · si PCR!D3	verified

Process

Protocol to delivery in about eight weeks

Scope the protocol

Assays, cell models, perturbations, replicates, readouts. We confirm what the lab can run and what the data will and will not support.

1 to 2 weeks

Run in a contract lab

A partner wet lab executes the protocol and keeps a full lab record: reagents with catalogue numbers, instrument settings, raw exports.

About 8 weeks per study

Extract and hash

Every value extracted with its cell reference kept. All source files hashed and registered. Annotations built as a separate layer.

1 week

Deliver and accept

JSONL, CSV, Parquet with schema, validator and dataset card. Acceptance checks on integrity, schema and source-cell consistency.

On delivery

FAQ

What buyers ask first

Is this enough data to train on?

No. The pilot is 330 rows from one study. Use it to test ingestion, schema and provenance. Volume comes from commissioned studies, delivered in the same format.

Who owns the data, and what can we do with it?

Rights documentation is confirmed per deal before any transfer. The free preview is for internal evaluation under a short agreement. Training rights, derivatives, exclusivity and duration are scoped per package.

Are there human samples involved?

The pilot uses primary human cells and donor-derived material. Ethics and privacy review status is stated per file in the registry and confirmed before transfer.

How do you verify the values?

Each value carries its workbook, worksheet and cell. We re-open the raw workbooks and compare before every delivery. For the preview, 12 of 12 rows matched on 4 November 2026.

Can you run our protocol?

Yes, for the six assay families above and standard adherent or suspension cell models. Send the protocol; we return a scope, a timeline and a quote.

What about expert annotation?

Captions and task references ship as machine-generated candidates. Expert review, band and lane annotation, gating and curve fits are a separately scoped service.

Contact

Request the free preview

30 measurement records, one Western blot image, ten flow-cytometry events, the data dictionary and sample tasks. Sent under a short evaluation agreement.

Mailbox being set up; replies may take a few days.

Or write tobd@orangeumbrella.io