cschanhniem · Jun 29, 2026 · Jun 29, 2026 · Jun 29, 2026
diff --git a/Makefile b/Makefile
@@ -1,4 +1,4 @@
-.PHONY: help demo test lint ci clean bench-leakage bench-baseline bench-hidden-active generate phase3 pilot validate-scoring validate-scoring-phase3 validate-scoring-strict external-predict pilot-confident presynth-qc gold-standard diversity synthesis-order novelty-broad
+.PHONY: help demo test lint ci clean bench-leakage bench-baseline bench-hidden-active generate phase3 pilot validate-scoring validate-scoring-phase3 validate-scoring-strict external-predict pilot-confident presynth-qc gold-standard diversity synthesis-order novelty-broad external-consensus questionnaire gate-check ip-report benchmark-card
 
 PYTHON := $(shell [ -f .venv/bin/python ] && echo .venv/bin/python || echo python3)
 PYTEST  := $(shell [ -f .venv/bin/pytest ] && echo .venv/bin/pytest || echo pytest)
@@ -17,8 +17,13 @@ help:
 	@echo "  make diversity          Within-panel diversity report"
 	@echo "  make synthesis-order    Vendor-ready synthesis order CSV + checklist"
 	@echo "  make novelty-broad      Compare pilot panel against 72-AMP curated reference set"
+	@echo "  make external-consensus Aggregate external predictor results into consensus verdicts"
+	@echo "  make questionnaire      Generate pre-populated reviewer questionnaires"
+	@echo "  make gate-check         Run all pipeline decision gates (gates 1-7)"
+	@echo "  make ip-report          Generate IP and novelty claim strength report"
+	@echo "  make benchmark-card     Generate/update benchmark summary card"
 	@echo ""
-	@echo "  make validate-scoring         AUROC on 43 AMP + 44 background (pipeline config)"
+	@echo "  make validate-scoring         AUROC on 95 AMP + 96 decoy (pipeline config)"
 	@echo "  make validate-scoring-phase3  AUROC with phase3.yaml config"
 	@echo "  make validate-scoring-strict  AUROC with scrambled-decoy strict benchmark"
 	@echo "  make bench-leakage            Check for near-duplicates between candidates and refs"
@@ -135,6 +140,33 @@ external-predict: pilot
 		--out-fasta outputs/pilot_panel.fasta \
 		--out-checklist outputs/external_predict_checklist.md
 
+# ── Sprint 3-6 targets ─────────────────────────────────────────────────
+
+external-consensus:
+	PYTHONPATH=src $(PYTHON) -m openamp_foundry.cli external-consensus \
+		--pilot-csv outputs/pilot_panel.csv \
+		--results-csv $$(or $$(RESULTS),outputs/external_predict_results.csv) \
+		--out outputs/external_consensus_report.md
+
+questionnaire:
+	PYTHONPATH=src $(PYTHON) -m openamp_foundry.cli reviewer-questionnaire \
+		--panel-csv outputs/confident_panel.csv \
+		--out outputs/questionnaire
+
+gate-check:
+	PYTHONPATH=src $(PYTHON) -m openamp_foundry.cli gate-check \
+		--validation-json outputs/validate_scoring_report.json
+
+ip-report:
+	PYTHONPATH=src $(PYTHON) -m openamp_foundry.cli ip-report \
+		--panel-csv outputs/confident_panel.csv \
+		--novelty-csv outputs/novelty_audit_full.csv \
+		--out outputs/ip_report.md
+
+benchmark-card: validate-scoring
+	@echo "Benchmark card is a static document at docs/BENCHMARK_CARD.md"
+	@echo "Run 'make validate-scoring' to update outputs/validate_scoring_report.json"
+
 pilot-confident:
 	@if [ -z "$(KEEP)" ]; then \
 		echo "Usage: make pilot-confident KEEP=ID1,ID2,..."; \

diff --git a/docs/BENCHMARK_CARD.md b/docs/BENCHMARK_CARD.md
@@ -0,0 +1,70 @@
+# Benchmark Card — OpenAMP Foundry v0.5.x
+
+> **Purpose:** Single-page summary of benchmark methodology, data, and metrics.
+> **Last updated:** 2026-06-29 (PR #110 expanded benchmark)
+
+---
+
+## Identity
+
+| Field | Value |
+|-------|-------|
+| Pipeline version | v0.5.x |
+| Benchmark type | Retrospective AUROC (composition-matched decoys) |
+| Date | 2026-06-29 |
+| Config | `configs/pipeline.yaml` (primary), `configs/phase3.yaml` (synthesis gate) |
+| Commit | See `docs/METRICS_CURRENT.md` |
+
+## Data
+
+| Set | Count | Description |
+|-----|:-----:|-------------|
+| Positives | **95** | Public-domain AMPs from 12 taxonomic classes |
+| Negatives (standard) | **96** | Length-matched random decoys (Swiss-Prot residue frequencies) |
+| Negatives (strict) | **95** | Per-sequence composition-matched shuffles |
+| Total (standard) | **191** | Expanded from original 87 (PR #110) |
+| Reference library | **120** | Unified AMP library (deduplicated curated_72 + expanded_95) |
+
+## Metrics
+
+| Metric | Standard (pipeline) | Phase3 gate | Strict |
+|--------|:-------------------:|:-----------:|:------:|
+| **AUROC** | **0.7832** | **0.7448** | 0.4335 |
+| **CI₉₅** | 0.72–0.84 | 0.68–0.81 | — |
+| **AUPRC** | 0.8164 | 0.7933 | — |
+| Baseline AUPRC | 0.4974 | 0.4974 | — |
+| Recall@10 | 0.1053 | 0.1053 | — |
+| Recall@20 | 0.2105 | 0.2105 | — |
+| Recall@43 | 0.4211 | 0.4000 | — |
+| Bootstrap | 2000 resamples | 2000 resamples | — |
+| **Interpretation** | **STRONG** | **STRONG** | Below random (expected) |
+
+## Method
+
+AUROC computed via Wilcoxon-Mann-Whitney statistic (concordant-pair enumeration).
+AUPRC via trapezoidal integration of precision-recall curve (pessimistic tie-breaking).
+Confidence intervals: percentile bootstrap (2000 resamples, seed=0).
+
+## Known Biases
+
+| Bias | Impact | Mitigation |
+|------|--------|------------|
+| Helical-centric scorer | β-sheet AMPs (defensins) score below panel | Noted in METRICS_CURRENT.md |
+| Melittin safety blind spot | Safety=1.0 despite strong hemolysis | Hemolysis assay mandatory for all |
+| Composition-matched scrambled AUROC < 0.5 | Model relies on composition > order | Expected for physicochemical scorer |
+| Near-seed generation only | Novel sequence space not explored | Acknowledged limitation in all docs |
+
+## Historical Baseline
+
+| Point | Set | Pipeline AUROC | Phase3 AUROC |
+|-------|:---:|:--------------:|:------------:|
+| Current (PR #110) | 95+96 (n=191) | **0.7832** | 0.7448 |
+| Pre-expansion (PR #72) | 43+44 (n=87) | 0.8420 | 0.8266 |
+| Pre-face-bonus (PR #70) | 43+44 | 0.8348 | 0.8126 |
+| Pre-windowed-mu_h (PR #66) | 43+44 | 0.8047 | 0.7846 |
+
+## Change Log
+
+| Date | Change | Author |
+|------|--------|--------|
+| 2026-06-29 | Initial card — expanded benchmark (PR #110) | OpenAMP CI |
diff --git a/docs/DECISION_RULES.md b/docs/DECISION_RULES.md
@@ -0,0 +1,90 @@
+# Pipeline Decision Rules
+
+> **Purpose:** Pre-registered pass/fail gates for each stage of the OpenAMP Foundry
+> pipeline. Thresholds are hardcoded (see `src/openamp_foundry/gates/gate_checker.py`)
+> to prevent cherry-picking after inspecting results.
+>
+> **Locked:** 2026-06-29 (Sprint 5)
+
+---
+
+## Gate 1 — AUROC Benchmark
+
+| Field | Value |
+|-------|-------|
+| Threshold | AUROC ≥ **0.70** |
+| Measure | Pipeline AUROC on expanded 95+96 benchmark |
+| Config | `configs/pipeline.yaml` |
+| Action | If FAIL: Do not proceed to synthesis |
+
+## Gate 2 — Leakage Guard
+
+| Field | Value |
+|-------|-------|
+| Threshold | Recall@43 < **0.60** |
+| Measure | Fraction of positives recovered in top 43 ranked |
+| Rationale | Recall@43 > 0.60 suggests near-duplicate memorisation |
+| Action | If FAIL: Investigate reference-set contamination |
+
+## Gate 3 — Model Disagreement
+
+| Field | Value |
+|-------|-------|
+| Threshold | Top-3 candidate |activity − boman_activity| < **0.45** |
+| Measure | Absolute difference between two independent scorers |
+| Config | `configs/pipeline.yaml`, `configs/phase3.yaml` |
+| Action | If FAIL: Review scorer weights or threshold |
+
+## Gate 4 — Top-10 Recall
+
+| Field | Value |
+|-------|-------|
+| Threshold | Recall@10 > **0.0** |
+| Measure | At least 1 positive in top 10 ranked candidates |
+| Action | If FAIL: Scoring model likely broken |
+
+## Gate 5 — Interpretation
+
+| Field | Value |
+|-------|-------|
+| Threshold | Benchmark interpretation must be **STRONG** |
+| Measure | String match of `interpretation` field |
+| Action | If FAIL: Do not proceed |
+
+## Gate 6 — External Predictor Consensus
+
+| Field | Value |
+|-------|-------|
+| Threshold | ≥3/5 external tools positive |
+| Measure | Manual web submission results |
+| Status | PENDING (see `outputs/external_predict_checklist.md`) |
+| Action | If FAIL: Expert reviewer override required |
+
+## Gate 7 — Human Expert Review
+
+| Field | Value |
+|-------|-------|
+| Threshold | Expert reviewer APPROVE or CONDITIONAL |
+| Measure | Completed reviewer questionnaire |
+| Status | PENDING (see `outputs/questionnaire/`) |
+| Action | If REJECT: Exclude from synthesis |
+
+---
+
+## CLI Usage
+
+```bash
+# Check all gates:
+openamp-foundry gate-check \
+    --validation-json outputs/validate_scoring_report.json
+
+# Check specific gate:
+openamp-foundry gate-check --gate 1 \
+    --validation-json outputs/validate_scoring_report.json
+```
+
+## Hardcoding Rule
+
+All gate thresholds in `gates/gate_checker.py` are **hardcoded** (not configurable
+via YAML). To change a threshold: modify source code, commit, and create a new release.
+This prevents silent threshold drift that could mask performance regression.
diff --git a/docs/EXTERNAL_PREDICTOR_CONSENSUS.md b/docs/EXTERNAL_PREDICTOR_CONSENSUS.md
@@ -0,0 +1,49 @@
+# External Predictor Consensus
+
+## Purpose
+
+Aggregate predictions from 5 independent AMP prediction tools into a single
+consensus verdict per candidate. A candidate with ≥3/5 tool agreement is
+classified as **CONFIDENT** and recommended for synthesis.
+
+## Tool Summary
+
+| Tool | Method | Positive label | URL |
+|------|--------|---------------|-----|
+| CAMPR4 | SVM + RF + ANN + DT ensemble | AMP | camp3.bicnirrh.res.in |
+| AMPScanner v2 | LSTM | Antimicrobial | dveltri.com/ascan/v2 |
+| dbAMP 2.0 | Random Forest | AMP | awi.cuhk.edu.cn/dbAMP |
+| AntiCP 2.0 | SVM | ACP (not AMP) | webs.iiitd.edu.in/raghava/anticp2 |
+| Macrel | Random Forest | AMP | big-data-biology.org/software/macrel |
+
+## Consensus Rules
+
+| Agreement | Verdict | Action |
+|:---------:|---------|--------|
+| ≥3/5 | CONFIDENT | Proceed to synthesis |
+| 2/5 | UNCERTAIN | Expert review required |
+| ≤1/5 | WEAK | Do not synthesise |
+
+## Current Status
+
+**PENDING** — Results not yet obtained. All 5 tools require manual web submission.
+See `outputs/external_predict_checklist.md` for the submission guide.
+
+## CLI Usage
+
+```bash
+# After filling in results CSV:
+make external-consensus RESULTS=outputs/external_predict_results.csv
+
+# Or manually:
+openamp-foundry external-consensus \
+    --pilot-csv outputs/pilot_panel.csv \
+    --results-csv outputs/external_predict_results.csv \
+    --out outputs/external_consensus_report.md
+```
+
+## See Also
+
+- `outputs/external_predict_checklist.md` — Manual submission guide
+- `src/openamp_foundry/reports/external_consensus.py` — Implementation
+- `docs/ASSAY_PREREGISTRATION.md` — Pre-registered assay protocol
diff --git a/docs/IP_PLAN.md b/docs/IP_PLAN.md
@@ -0,0 +1,79 @@
+# IP and Disclosure Plan
+
+> **Purpose:** Strategic guidance for intellectual property decisions before public
+> sequence disclosure. Not a legal opinion — consult a patent attorney before filing.
+>
+> **Last updated:** 2026-06-29 (Sprint 6)
+
+---
+
+## Patent Strategy
+
+### Claimable Candidates
+
+| Seed | Classification | Claim strategy |
+|------|---------------|----------------|
+| SEED-006 (4 candidates) | HIGH_CONFIDENCE_NOVEL | Primary patent claims — novel wasp-venom scaffold |
+| SEED-007 (1 candidate) | HIGH_CONFIDENCE_NOVEL | Primary patent claims — novel bumblebee-venom scaffold |
+| SEED-008 (3 candidates) | HIGH_CONFIDENCE_NOVEL | Primary patent claims — novel Trp-rich plant scaffold |
+| SEED-009 (2 candidates) | HIGH_CONFIDENCE_NOVEL | Primary patent claims — novel proline-rich scaffold |
+| SEED-007 (3 candidates) | NOVEL | Moderate claims — verify mechanism distinct from bombolitin literature |
+| SEED-008 (1 candidate) | NOVEL | Moderate claims — verify mechanism distinct from indolicidin |
+| SEED-005 (1 candidate) | CLOSE_RELATIVE | Not individually patentable — SAR control |
+| SEED-001 (1 candidate) | KNOWN_VARIANT | Not patentable — positive control |
+| SEED-003 (2 candidates) | KNOWN_VARIANT | Not patentable — tachyplesin-like SAR |
+
+### Provisional Patent
+
+Before any public sequence disclosure, file provisional patent application covering:
+1. SEED-006 scaffold and all pilot variants
+2. SEED-007 scaffold and all pilot variants
+3. SEED-008 scaffold and all pilot variants
+4. SEED-009 scaffold and all pilot variants
+5. Composition claims for any MIC ≤ 8 µg/mL candidate
+6. Methods of use (antimicrobial treatment, wound healing)
+
+### Sequence Deposit
+
+Deposit sequences in patent filing as SEQ ID NOs. Use the WIPO ST.26 standard
+(valid from July 2022) for sequence listing.
+
+## Disclosure Plan
+
+| Tier | Candidates | Disclosure | Timing |
+|:----:|------------|------------|--------|
+| 1 | SEED-006, 007, 008, 009 (14 candidates) | **Redacted** from public repo | After provisional filing |
+| 2 | SEED-001, 003, 005 (6 candidates) | Public (known scaffolds) | Anytime |
+| 3 | Pipeline code | Public (MIT license) | Already public |
+
+## Freedom to Operate
+
+**Required searches before publication:**
+- [ ] APD3 BLASTp — no hit > 80% identity
+- [ ] DRAMP patent section — no overlapping claims
+- [ ] Google Patents / Lens.org — scaffold-level search
+- [ ] Competitor sequence database (AMP-Designer, AMPGAN v3, etc.)
+
+## Publication Strategy
+
+1. **First publication:** Pipeline methodology + benchmark results only
+   (no candidate sequences disclosed)
+2. **Second publication:** Candidate nominations + novelty audit
+   (after provisional patent filing)
+3. **Third publication:** Wet-lab results
+   (after independent replication)
+
+## Material Transfer
+
+Before sharing candidate peptides with external labs, execute an MTA covering:
+- Sequence confidentiality
+- Publication rights
+- IP ownership (background vs foreground)
+- Data sharing obligations
+
+## Contributor IP
+
+All contributors to this repository have assigned code contributions under
+the MIT license. No separate contributor agreement is in place for computational
+contributions. If wet-lab collaborators join, a formal collaboration agreement
+should define IP ownership before any experimental work begins.