| name | indication-dossier |
| description | Build an evidence-traceable indication dossier for a disease, target, mechanism, drug, biomarker, or therapeutic hypothesis. Use for landscape assessment, target-indication rationale, clinical pipeline review, translational gaps, and evidence-backed go/no-go questions; not for patient-specific advice. |
| license | MIT |
Indication Dossier
Scope
Define the indication, population, stage, geography, intervention/target,
decision, evidence cutoff, and non-goals before searching. Separate scientific,
clinical, regulatory, commercial, and operational claims.
Workflow
- Resolve disease/target/drug identifiers with OLS, UniProt, Open Targets,
ChEMBL, PubChem, and FDA sources. Preserve synonyms and ontology IDs.
- Search primary literature with PubMed, Europe PMC, bioRxiv, and OpenAlex;
search registered studies with ClinicalTrials.gov. Record exact queries,
dates, inclusion/exclusion criteria, and deduplication.
- Build evidence tables for disease biology, genetics, expression/cell context,
target validation, mechanism, preclinical models, biomarkers, safety/liability,
clinical competitors, trial status/results, regulatory facts, and major gaps.
- Grade each claim by evidence type, independence, directness, recency, model
relevance, replication, and conflict. Distinguish registered, completed,
reported, peer-reviewed, approved, terminated, and merely announced states.
- Produce an executive summary, evidence map, pipeline table, contradictions,
uncertainties, falsifiable next experiments, and explicit go/no-go criteria.
- Save citations, tables, queries, snapshots, and claim links under
artifacts/<run-id>/indication-dossier/ with $science-provenance; run
$science-review before delivery.
Boundaries
- Do not infer efficacy from mechanism, preclinical evidence, or trial existence.
- Do not provide diagnosis, treatment, dosing, or patient-matching advice.
- Clearly label inference, missing results, stale records, and conflicts of interest.
