Every experiment,
accounted for.

ChemCellar is an open-source platform for registering and managing compounds, analysing screening campaigns and deciding what to make next, self-hosted on infrastructure you control.

Self-host it View on GitHub

Registration
Search
Protocols
Screening
SAR workbench
Campaigns
Inventory

Open sourceAGPL-3.0 license, public repository.

PostgreSQL + RDKitChemistry lives in the database.

Docker ComposePublished images, for the release you choose.

FIG. 1Run R-0142 | EGFR kinase | 384-well | 10-point dose responseZ′ 0.81

CC-004217BATCH 02 | I13–J22 | n = 2

MW446.91

cLogP4.28

IC₅₀33 nM

Hill slope1.05

One record, from registration to decision.

The structure you register is the same record your assay results, SAR tables and freezer locations point to. Nothing is copied between systems, so nothing drifts.

1CompoundCC-004217

(i)register

2BatchB02 | 12.4 mg | −20 °C

(ii)plate

3RunR-0142 | 384-well

(iii)fit

4ResultIC₅₀ 33 nM | Z′ 0.81

(iv)decide

5DecisionStage: Lead

Reagents and conditions: (i) standardize, strip salts, check for duplicates; (ii) weigh, barcode, dispense at ten concentrations; (iii) four-parameter fit, Z′ ≥ 0.5; (iv) stage criteria, reviewed by the team.

The same molecule is registered once.

Draw it, paste it or upload a file. ChemCellar standardizes the structure, records the salt, and, if the compound is already in the cellar, adds your material as a new batch of it.

Register compoundInput / Preview / Batch / Summary

Project: EGFR-2

As drawn

COc1cc2ncnc(Nc3ccc(F)c(Cl)c3)c2cc1OCCCN1CCOCC1.Cl

Registered parent

Salt stripped: HCl, recorded on the batch
Charges neutralized, parent extracted
InChIKey XGALLCVXEZPNRQ-UHFFFAOYSA-N

This structure is already registeredCC-004217 / registered 2 Mar 2026 / 1 batch

Add as batch 02

STANDARDIZEOne canonical parentSalts and charges handled by the ChEMBL structure pipeline.
DEDUPLICATEStereo-aware identityMatched on InChIKey: enantiomers stay distinct, common tautomers are treated as one.
BULKFrom one compound to a libraryImport SDF, CSV or Excel with a preview before anything is written.

Ask by structure. Filter by result.

Combine a substructure with assay criteria, properties, tags and projects in one query. Every value in the answer shows how many runs stand behind it.

STRUCTURESubstructure, similarity, exactSubstructure and similarity run in PostgreSQL through the RDKit cartridge; exact match is by InChIKey.
RESULTSCriteria across any protocol“Active anywhere below 1 µM” is one condition, not twenty.
HONEST NUMBERSYou choose how runs combineLatest run, geometric mean or best fit by R², with disagreements flagged.
REUSESaved, shared, exportedCSV, SDF, Excel with numbers stored as numbers, or a PDF report.

Search/ EGFR-2 / Results

Runs: Geometric mean ▾Export ▾Save search

SubstructureEGFR IC₅₀ ≤ 500 nMAny protocol / activeMW ≤ 600Tag: series-A6 of 1,284 / 0.21 s

ID	MW	EGFR IC₅₀ (nM)	Active in	Stage
CC-004302Afatinib	485.95	0.5n=4	EGFR, HER2 +2	Lead
CC-004188Erlotinib	393.44	2.0n=6	EGFR, A431	Lead
CC-004415Dacomitinib	469.95	6.0n=3⚠	EGFR, HER2 +1	Confirmed
CC-004217Gefitinib	446.91	33n=2	EGFR, A431	Confirmed

Structure

EGFR IC₅₀ (nM)

Curve

Active in

Stage

CC-004302Afatinib

485.95

0.5n=4

EGFR, HER2 +2

Lead

CC-004188Erlotinib

393.44

2.0n=6

EGFR, A431

Lead

CC-004415Dacomitinib

469.95

6.0n=3⚠

EGFR, HER2 +1

Confirmed

CC-004217Gefitinib

446.91

33n=2

EGFR, A431

Confirmed

An assay catalog that stays navigable.

Assay lists rot: near-identical protocols, run details baked into names, five spellings of one category. ChemCellar is designed so that does not happen, without telling anyone how to name things.

Protocols/ Library

Group by: Target ▾New protocol

EGFR12 protocols

EGFR kinase, ADP-GloBiochemical / ADP-Glo / IC₅₀, % inhibition

42 runs

Z′ 0.79

EGFR L858R/T790M kinaseBiochemical / HTRF / IC₅₀

17 runs

Z′ 0.74

A431 proliferationCell-based / CellTiter-Glo / GI₅₀

23 runs

Z′ 0.68

HER27 protocols

HER2 kinase, ADP-GloBiochemical / ADP-Glo / IC₅₀

31 runs

Z′ 0.77

BT-474 proliferationCell-based / CellTiter-Glo / GI₅₀

12 runs

Z′ 0.66

This looks like a run of an existing method

“EGFR kinase ADPGlo 10pt” shares its readouts and its target with EGFR kinase, ADP-Glo (v3).

Log a run of thisDismiss

REUSESuggested, never blockedWhen a new protocol resembles an existing one, you are shown it and you decide.
NAVIGATEFacets with live countsSlice by target, format, detection, organism or status.
VERSIONMethods change, history staysReadouts and conditions are versioned with the protocol.

From plate reader to fitted curve.

Import a run, check the plate, fit the curves, lock the data. Every value keeps its link to the well, the batch and the compound.

IMPORTOne import, any layoutPlate files or summary tables. The original file is kept with the run.
QUALITYZ′ and controls on every plateSee a bad plate before it becomes a bad decision.
FITCurves you can inspect and correctExclude a point, refit, and keep the history of why.
LOCKHit criteria with an authorWho set the threshold, and when, is part of the record.

R-0142EGFR kinase, ADP-Glo / 384-well / 14 Aug 2026

Z′ 0.81Locked

CC-004217Wells I13–J22 / n = 2

IC₅₀33 nM

Hill slope1.05

Top98.4

R²0.994

See which change mattered.

Map a whole library by chemical similarity, colour it by potency, and draw around the region that interests you. This map is live: drag across it to make your own selection.

Cluster mapEGFR-2 library | 1,920 compounds | Morgan fingerprints

Drag to draw your own lasso

pIC50 59

SCAFFOLD TREESeries, organised for youCompounds grouped by shared ring systems, at library scale.

R-GROUP TABLEEvery substituent against every assayPick a core, and the series lines up as a heatmap.

CLUSTER MAPDraw around what you wantLasso a region of chemical space and save it as a collection.

Every promotion has a reason.

A campaign pulls results from the protocols you choose and walks compounds through named stages. Criteria do the routine work. People make the calls, and the call is recorded.

EGFR-2 / Primary to lead5 channels / 3 libraries

OpenPreview as published

1,920

Screened

3 libraries

214

Primary hit

≥ 50 % at 10 µM

Confirmed

IC₅₀ ≤ 1 µM

Selective

HER2 / EGFR ≥ 10×

Lead

Reviewed

ID	EGFR IC₅₀	HER2 IC₅₀	Selectivity	A431 GI₅₀	Stage
CC-004188	2.0 nM	> 10 µM	> 5,000×	110 nM	Lead
CC-004217	33 nM	3.7 µM	112×	80 nM	Leadpromoted
CC-004415	6.0 nM	46 nM	7.7×	35 nM	Confirmed
CC-004530	500 nM	> 10 µM	> 20×	ND	Primary hit

EGFR IC₅₀

HER2 IC₅₀

Selectivity

A431 GI₅₀

Stage

CC-004188

2.0 nM

> 10 µM

> 5,000×

110 nM

Lead

CC-004217

33 nM

3.7 µM

112×

80 nM

Leadpromoted

CC-004415

6.0 nM

46 nM

7.7×

35 nM

Confirmed

CC-004530

500 nM

> 10 µM

> 20×

Primary hit

OverrideCC-004217 promoted to Lead: “Potency confirmed in the orthogonal binding assay.”

STAGESA funnel you defineName the stages and set the criteria that move a compound forward.
JUDGEMENTPromote or demote, with a reasonOverrides are allowed, visible, and attributed.
OUT OF RANGE“> 10 µM” is not 10 µMA value that cannot prove a criterion does not pass it.

Know where every milligram is.

Batches, samples and plates are tracked from synthesis to the freezer box, and out again when they are loaned, requested or shipped.

LOCATEFreezer, rack, box, positionBrowse storage the way it is physically laid out.
MOVELoans, requests and shipmentsChain of custody from your bench to a partner lab.
MAKESynthesis requestsAsk for more, assign it, and follow it to a new batch.
SCANKiosk mode at the benchBarcode stations for check-in and check-out.

Inventory/ Storage / Box 14

58 of 81 positions filled

Location

Freezer 2 (−20 °C) / Rack B / Box 14 / C7

CC-004217-B02C7

Amount: 12.4 mg of 20.0
Purity: 99.1 % (LC-MS)
Salt: HCl, 1 eq
Source: Synthesis SR-0088

Loan2.0 mg to Assay Labdue 14 Oct
ShipmentSH-0031 to partner sitein transit
RequestSR-0102 resynthesis, 50 mgassigned

The details are the product.

Most of what makes discovery data trustworthy is small. These are decisions ChemCellar makes the way a careful scientist would.

A “>” stays a “>”

A result reported as “greater than” keeps its qualifier through every table, criterion and export.

Potency averages on the log scale

Geometric means and fold-range across runs, with a flag when runs disagree by more than tenfold.

Identity is stereo-aware

Duplicates are matched on InChIKey. Enantiomers are never merged into one compound.

The raw file stays with the run

Every import keeps the original file, so a curve can be defended months later.

Thresholds have an author

Hit criteria record who chose them and when, and freeze when the run is locked.

Nothing disappears quietly

Changes land in an audit trail that cannot be edited. Deletes show what they will affect first.

Your data never leaves the building.

Discovery data is the project. ChemCellar is built so the group doing the science is the group holding the keys.

OpenEvery line of code is public. Read it, audit it, fork it.

YoursRuns on your hardware, with sign-in through your own identity provider.

PortableExport everything, any time, in open formats.

How self-hosting works →

$ git clone https://github.com/sidxz/cellar
$ cd cellar && cp .env.example .env
  # point it at Duar and your identity provider
$ make prod-up

✔ postgres   RDKit cartridge      healthy
✔ migrate    schema at head       done
✔ backend    :8000                ready
✔ frontend   :3000                ready

Works withDocuStore

Chemistry-aware document intelligence. Compounds and bioactivity extracted from your reports and papers.

Works withDAIKON

Project and pipeline tracking for discovery programs.

Try it with real chemistry.

Self-host it View on GitHub

Every experiment,
accounted for.

One record, from registration to decision.

The same molecule is registered once.

Ask by structure. Filter by result.

An assay catalog that stays navigable.

Target

Assay format

Detection

Status

From plate reader to fitted curve.

See which change mattered.

Every promotion has a reason.

Know where every milligram is.

The details are the product.

A “>” stays a “>”

Potency averages on the log scale

Identity is stereo-aware

The raw file stays with the run

Thresholds have an author

Nothing disappears quietly

Your data never leaves the building.

Try it with real chemistry.

Every experiment,accounted for.

One record, from registration to decision.

The same molecule is registered once.

Ask by structure. Filter by result.

An assay catalog that stays navigable.

Target

Assay format

Detection

Status

From plate reader to fitted curve.

See which change mattered.

Every promotion has a reason.

Know where every milligram is.

The details are the product.

A “>” stays a “>”

Potency averages on the log scale

Identity is stereo-aware

The raw file stays with the run

Thresholds have an author

Nothing disappears quietly

Your data never leaves the building.

Try it with real chemistry.

Every experiment,
accounted for.