ChEMBL Bioactivity Normalizer & QC

Turn official ChEMBL activity records into source-faithful, analysis-ready measurement rows with conservative molar normalization, explicit relation semantics, QC flags, provenance, and stable SHA-256 record hashes.

This Actor is for medicinal-chemistry, cheminformatics, structural-biology, and data-curation workflows that need to know what a measurement actually says before comparing it. It does not rank compounds, infer potency, or turn censored bounds into exact values.

What you receive

Every admitted dataset row includes:

• original and standardized type, relation, value, units, and text value exactly as supplied by ChEMBL;
• activity, assay, molecule, target, and document identifiers;
• a molar value only when the standard numeric value, unit, and relation can be interpreted safely;
• an explicit relation class (exact, upper_bound, lower_bound, approximate, or unknown_relation);
• a conservative comparability class and human-review grouping key;
• sorted QC flags and exclusion reasons, including ChEMBL validity comments without suppression;
• a deterministic record hash that ignores only retrieval/run metadata;
• record-level ChEMBL source, retrieval, attribution, license, and ShareAlike fields.

Unsupported measurements are still useful QC outcomes. For example, a percent-inhibition record is returned as not_molar_comparable with an unsupported_unit flag; it is never silently converted or discarded.

Input

Supply activity IDs, one constrained query, or both. The Actor accepts no arbitrary URL, header, cookie, credential, webhook, file, or executable input.

Direct records:

{
  "activityIds": [31863],
  "maxConcurrency": 2,
  "requestTimeoutSeconds": 20,
  "maxRetries": 3
}

Bounded query:

{
  "query": {
    "targetChemblId": "CHEMBL204",
    "standardType": "IC50",
    "maxRecords": 100
  }
}

Query filters are exact ChEMBL IDs for a molecule, target, assay, or document, plus an optional standard type. maxRecords is required and capped at 5,000. Concurrency is capped at 3 and all requests go only to the official HTTPS ChEMBL API.

Output example

{
  "schemaVersion": "chembl-normalized-activity/v1",
  "activityId": 31863,
  "canonicalMolarValue": 0.0001,
  "relationClass": "lower_bound",
  "comparabilityClass": "normalized_censored",
  "qcFlags": ["censored_relation"],
  "exclusionReasons": [],
  "sourceName": "ChEMBL",
  "sourceLicense": "CC BY-SA 3.0",
  "recordHash": "64-lowercase-hex-characters",
  "retrievedAt": "UTC RFC 3339 timestamp",
  "runId": "non-secret run identifier"
}

The actual row also contains the complete documented identifier, measurement, context, validity, attribution, source URL, and ShareAlike fields. A final aggregate summary is written uncharged to the RUN_SUMMARY key-value record rather than the paid dataset. It reports rows written and events charged separately, so local non-PPE testing remains truthful.

Normalization rules

Only exact M, mM, uM, µM, μM, nM, and pM tokens are recognized as molar concentration units. The original unit text is always retained. A canonical molar value is emitted only when the standard value is finite and strictly positive and its relation is one of =, <, <=, >, >=, or ~.

Source relation	Meaning preserved	Comparability class when molar
=	exact	normalized_exact
<, <=	upper bound	normalized_censored
>, >=	lower bound	normalized_censored
~	approximate	normalized_approximate
missing/other	unknown	not_molar_comparable

The grouping key means only “eligible for human review together.” Assays can differ for many scientifically important reasons. This Actor never aggregates, averages, imputes, ranks, or claims interchangeability.

Pay per event

• normalized-activity: $0.006 per successfully written result row.
• Apify supplies the separate synthetic Actor-start event.
• Source failures, invalid inputs, malformed/unidentifiable source records, and work stopped before output are not charged a row event.
• A valid but unsupported measurement is a paid QC transformation because it preserves the source record and explains why it should not be treated as molar-comparable.
• The Actor checks the run’s charge limit before starting or admitting more work and stops with a bounded summary when the limit is reached.
• Platform usage is included in the pay-per-event price.

Source, license, and attribution

Data is retrieved only from the official ChEMBL activity API. Every output row attributes ChEMBL, links the source record, records retrieval time, identifies CC BY-SA 3.0, links the license, and states that derived output remains subject to attribution and ShareAlike. The Actor does not fetch or republish commercially calculated molecule-property fields or third-party structures.

Unofficial transformation tool; not affiliated with or endorsed by ChEMBL or EMBL-EBI.

ChEMBL-derived output remains subject to CC BY-SA 3.0. Review ChEMBL’s current terms and attribution guidance for your downstream use.

Privacy and security

The Actor has Limited permissions, no Standby server, no arbitrary network destinations, and no credential-bearing input. It does not put raw HTTP bodies, response headers, environment values, machine/user names, local paths, tokens, or cookies in results. Inputs, normalized rows, and the run summary use Apify storage according to the retention and access settings of your Apify account.

Limits

• This is deterministic transformation and QC, not medicinal, clinical, safety, efficacy, or treatment advice.
• A normalized value does not prove that two assays are scientifically interchangeable.
• Source outages can yield a truthful partial run. The summary reports source failures and partial coverage.
• Upstream ChEMBL corrections can change deterministic fields and therefore the record hash in a later run.
• v0.1 does not fetch structures, calculated molecular properties, abstracts, or third-party enrichment.

Support

Flintglade · support@flintglade.com