Brown Biotech AI drug discovery research workflow. PaperOrchestra-enhanced pipeline: target discovery → virtual screening → experiments → paper writing with Semantic Scholar citation verification. Use when user says 'Brown Biotech 연구', 'drug discovery paper', '신약 연구 파이프라인', or wants end-to-end biotech research.
Research direction: $ARGUMENTS
PaperOrchestra-enhanced research pipeline optimized for AI drug discovery. Combines ARIS autonomous workflows with PaperOrchestra's structured planning and citation verification.
Phase 1: Target Discovery /idea-discovery (biotech-adapted)
↓
Phase 2: Literature + Prior Art /research-lit + /semantic-scholar (dual-source)
↓
Phase 3: Experiment Design /experiment-plan (claim-driven)
↓
Phase 4: Virtual Screening /experiment-bridge + /run-experiment
↓
Phase 5: Paper Writing /paper-writing (Orchestra-enhanced)
↓
Phase 6: Cross-Model Review /auto-review-loop (difficulty: hard)
↓
Phase 7: Submission + Rebuttal /rebuttal (if needed)
Decompose into: Disease/Target, Method, Data, Gap.
Run /idea-discovery "$ARGUMENTS" with biotech search strategy:
Run /novelty-check against arXiv + Semantic Scholar.
Gate 1: Novelty score >= 7/10.
/research-lit "$ARGUMENTS" — ARXIV_DOWNLOAD: true
/semantic-scholar "$ARGUMENTS" - fields: all, min-citations: 10, year: 2023-
Build structured map: Core → Method → Dataset → Baseline → Application papers. Gate 2: >= 30 verified citations.
Define testable claims via /experiment-plan:
Run /ablation-planner. Ensure seeds, splits, hyperparams documented.
Gate 3: Plan approved.
/experiment-bridge → /run-experiment
/monitor-experiment + /training-check
/analyze-results → /result-to-claim
Gate 4: >= 1 primary claim supported.
/paper-plan with visualization plan + citation hints per section.
/paper-figure — conceptual diagrams + statistical plots.
/paper-write with CITATION_VERIFY=true:
Every citation verified via Semantic Scholar API. Hallucinated → [VERIFY].
/paper-compile
Gate 5: PDF compiles, all citations verified.
/auto-review-loop — difficulty: hard, max rounds: 3
Gate 6: Score >= 7/10.
Final checklist + /rebuttal if needed.
virtual screening, molecular docking, pharmacophore, QSAR, ADMET, drug-likeness, Lipinski, protein-ligand, binding affinity, molecular generation, de novo design, GNN, transformer, diffusion, hit-to-lead, lead optimization
project/
├── literature/ # Survey results + citation_graph.json
├── experiments/ # Code, results, EXPERIMENT_PLAN.md
├── paper/ # LaTeX + PDF
└── RESEARCH_LOG.md