1. Publication brief
Define the scientific question, publication type, accountable product lane, and intended journal before retrieving evidence.
2. Protocol and evidence plan
Make the analysis reproducible before results are known. Registration may be marked not required only with an explicit rationale.
Advanced retrieval settings
3. Evidence library
Retrieve, deduplicate, verify, and monitor every source. Demonstration records are never silently treated as publication evidence.
4. Two-reviewer screening
AI may prioritize records, but final inclusion requires accountable reviewer decisions, conflict resolution, and explicit exclusion reasons.
5. Verified extraction
Record values with their source, endpoint, units, uncertainty and verification state. Never pool unverified placeholders.
Full-text / PK extraction assistant
6. Analysis and diagnostics
Run synthesis only on verified extraction rows. Preserve model choice, effect scale, software version, diagnostics and sensitivity results.
7. Bayes Pharma evidence
Import governed outputs as versioned EvidencePackets. Endpoint availability alone is not evidence and does not imply model validation.
8. Evidence-bound manuscript
Draft only from the project evidence graph. Verification placeholders and unsupported claims block submission readiness.
9. Reporting, authorship and review
Complete study-specific reporting guidance, CRediT roles, declarations, privacy requirements, and accountable approvals.
10. Journal submission package
Draft exports remain available throughout review. The final submission ZIP unlocks only after every scientific, reporting, privacy and approval gate passes.
Advanced capability laboratory
Existing publication APIs remain available for expert use. Outputs enter the governed project only after explicit verification.
{}