Wikidata Bindingdb Lookup

Skill workflow-skill many-tools-to-one-skill Proprietary (Paramus) — guidance only; execution performed by Paramus tools under their own license gating.
provenanceai generated statuspartial
Look up measured activity for a compound/target pair across the public activity databases and return a cited table - identifiers reconciled, units normalized, conflicts reported with sources.When to use: a user names a compound and a target and asks “what activities are published for this pair?”.Multi-tool WORKFLOW with identifier-reconciliation and unit-normalization judgment.
Procedure6 steps · 4 orchestrated tools
1
Discover tools via search/get_schema ("search compound", "compound info", "activity database").
2
Resolve the compound identity: cloud_pubchem_search on the name or structure,
Toolscloud_pubchem_search
3
Query the activity databases for the pair - in implementation (wikidata-chem),
4
Reconcile identifiers across the databases: map Wikidata item, PubChem CID and
5
Normalize the units: convert to a common unit for comparison, keep the raw value
6
Report the cited table: one row per measurement with database, value (raw and
You get
the identity resolution with identifiers per database; the cited activity

Overview

The same compound is a CID here, a Wikidata item there and a BindingDB monomer somewhere else, and its activity for one target is scattered across all three with different units. This skill answers "what is published for this compound/target pair" with a cited table: the pair resolved in each database, identifiers reconciled, units normalized, and conflicts reported with their sources instead of averaged away. The value is the reconciliation and the conflict honesty; the database apps are queued.

Guidance vs execution: identifier search and compound info come from tested Paramus tools. The activity-database lookups are in implementation (wikidata-chem), (bindingdb) and (comptox); do not scrape the databases in ad-hoc code.

When to Use

  • A compound and a target are named and the ask is "what measured activities are published for this pair, with citations?"
  • Reconciling why two sources report different numbers for the same pair

Do not use for: predicting activity that was never measured (a modelling task), or bulk mining of a whole target's ligands without a named compound. One pair, cited - that is the trigger.

Domain Judgment

  • Identity before numbers: a perfect IC50 for the wrong tautomer or salt is a wrong answer delivered confidently - reconcile the identifiers first.
  • Keep the raw value beside the normalized one: unit conversion is a judgment the reader must be able to audit.
  • A conflict between sources is a result, not a problem: report both values with their citations and let the reader weigh assay conditions.

Failure modes

  • The compound cannot be resolved to an identity -> stop and ask for a structure or a precise identifier; do not look up a guess.
  • Identity conflict between databases -> flag the pair as ambiguous and report both candidate identities; never merge on a coin flip.
  • The activity-database apps are unavailable -> deliver the identity reconciliation from the existing tools; mark the activity lookup open, invent no values.

Example

Asked: "What activities are published for imatinib against ABL1, with sources?" The skill resolves the identity with cloud_pubchem_search and cloud_pubchem_info, then queries the activity databases - in implementation (wikidata-chem), (bindingdb), (comptox). Reconciliation finds the compound under one PubChem CID, a Wikidata item and several BindingDB ids (free base and mesylate salt), and the table lists them as one compound with the salt forms noted. Unit normalization puts every value in nM for comparison while keeping each raw value and unit beside it. Where two sources differ by an order of magnitude, the conflict list reports both with citations and assay types rather than averaging them away.

Input

compound (name, SMILES or identifier), target (name or accession),

Output

the identity resolution with identifiers per database; the cited activity

Orchestrates · 4 tools

The skill decides at runtime which to call.

cloud_pubchem_search cloud_pubchem_info search_chemical_by_formula calculate_similarity