cschanhniem · Jun 29, 2026 · Jun 29, 2026 · Jun 29, 2026
diff --git a/docs/EXPERT_REVIEW_PACK.md b/docs/EXPERT_REVIEW_PACK.md
@@ -174,6 +174,19 @@ not a claim that they are inactive.
 | SEED-007 (bombolitin) | 4 | Short amphipathic helix | Genuine (✓ NOVEL) | Low (11 AA; good) | ~70–82% | Novel short helix |
 | SEED-008 (Trp-rich) | 4 | Indole ring stacking; not helical | Genuine (✓ NOVEL) | Moderate (13 AA model edge) | ~70–82% | Novel mechanism; high disagreement |
 | SEED-009 (Pro-rich) | 4 | Ribosome exit + DnaK | Genuine (✓ NOVEL) | **Best** (Pro-X bonds resist protease) | ~60–75% | Novel; run RPMI parallel |
+| SEED-019/020 (Arg-Ile/Val/Leu) | 3 | Cationic amphipathic helix | ✓ NOVEL (SEED-020 pair); CLOSE_RELATIVE + ⚠ patent risk (SEED-019_VAR_004) | Short (12 AA model edge) | **~75–85%** (SEED-020 pair); patent clearance needed for SEED-019_VAR_004 | Best ext.-predictor confirmed leads; IP check before SEED-019 synthesis |
+
+> **Wave 0.5b addendum (2026-06-29):** External predictor screening (CAMPR4 × 4 votes,
+> AMPScanner v2, Macrel, AMPActiPred, HemoFinder, AntiCP2) of 23 new SEED-020/021/023/024
+> candidates confirmed SEED-020_VAR_004 (RLRIRVLKRLLK) and SEED-020_VAR_002 (KVRIRVLKRLLK)
+> as the best safety-activity tradeoff in the batch: HemoFinder LOW, AMPScanner > 0.99,
+> Non-AntiCP. Both confirmed NOVEL by broad novelty check (72-AMP curated DB; < 50% max
+> similarity; no patent DB matches). Both added to `outputs/wave1_final_panel.csv` as
+> BALANCED_LEAD candidates (source: WAVE0_5B). SEED-019_VAR_004 (already in panel) carries
+> a POSSIBLE_PATENT_RISK flag (66.7% to DRAMP05502 in full 27,234-sequence audit) — require
+> IP clearance before synthesis. SEED-022 was excluded at pre-screen (no viable candidates
+> after safety scoring; poor AntiCP / HemoFinder profile across all variants). Full synthesis
+> notes: `docs/WET_LAB_HANDOFF.md §SEED-019/SEED-020`.
 
 **SEED-007 reviewer note:** Bombolitin-II (*Megabombus pennsylvanicus* venom) is the
 **largest selected family** (26 variants, 4 pilot slots). All 4 pilot candidates are NOVEL

diff --git a/docs/ROADMAP.md b/docs/ROADMAP.md
@@ -95,6 +95,8 @@ Implemented during the pre-wet-lab improvement loop (PRs #31–#54):
 - EXPERT_REVIEW_PACK.md: Pipeline Limitations table updated — PYROGLUTAMATE_RISK row added (N-terminal Q, physiological pH, none of 20 pilot candidates affected); SEED-009_VAR_033 Met row added (PR #108)
 - test: close remaining 1% coverage gap — 6 modules to 100% branch coverage; 1321 tests total; only 6 structural CLI guard lines remain uncovered; all source modules at 100% branch coverage (PR #109)
 - feat(bench+qc): expanded benchmark (95 AMPs + 96 decoys, n=191, AUROC=0.7832, CI₉₅: 0.72–0.84); 52 new public-domain AMPs across 12 classes (defensins, proline-rich, lantibiotics); DKP_RISK flag for N-terminal X-Pro diketopiperazine cyclization — all 4 SEED-008 pilots (F-Pro) flagged with dynamic mass (244 Da), MS receipt check, Nα-Ac REQUIRED; WET_LAB_HANDOFF.md SEED-008 synthesis guidance updated; ASSAY_PREREGISTRATION.md SEED-008 DKP note added; 1337 tests (PR #110)
+- fix(docs+qc): correct SEED-006 Met synthesis error — ALL 4 SEED-006 pilots contain Met at position 9 (inherited from mastoparan-X); Nle/M9Nle REQUIRED in synthesis order; ASSAY_PREREGISTRATION.md SEED-006 Met pre-registered; EXPERT_REVIEW_PACK.md: SHORT_PEPTIDE count corrected (6→10 candidates), DKP_RISK + Met rows added, AUROC stale values corrected; regression guard `test_seed006_all_pilots_have_met_at_position9` with positional seq[8]=='M' assertion in TestOxidationRisk (PR #104)
+- feat(panel): wave 0.5b external-predictor confirmation — SEED-020_VAR_004 (RLRIRVLKRLLK, AMPScanner 0.9928, HemoFinder LOW, Non-AntiCP 0.29) and SEED-020_VAR_002 (KVRIRVLKRLLK, AMPScanner 0.9960, HemoFinder LOW, Non-AntiCP 0.32) confirmed NOVEL by broad novelty check (72-AMP DB, <50% similarity); both added to wave1_final_panel.csv as BALANCED_LEAD (WAVE0_5B source); SEED-019/020 synthesis notes added to WET_LAB_HANDOFF.md; EXPERT_REVIEW_PACK.md §5 updated with SEED-019/020 family row and wave 0.5b addendum
 
 ## v0.5.7 — Novelty Audit v2: Proper Prior-Art Scan ✓ (2026-06-29)
 

diff --git a/docs/WET_LAB_HANDOFF.md b/docs/WET_LAB_HANDOFF.md
@@ -441,6 +441,76 @@ Pro-rich AMPs show > 2 h stability in 50% human serum (Otvos & Cudic 2002, *Curr
 
 ---
 
+## SEED-019/SEED-020 (Arg-Ile/Val/Leu Alternating Scaffold) Special Notes
+
+SEED-019 and SEED-020 form a tight SAR cluster of short cationic amphipathic helices sharing
+the alternating Arg-Ile/Val/Leu pattern. All candidates are 12 AA with 5–6 combined Lys/Arg
+residues. SEED-020 variants are design-inspired mutations of SEED-019_VAR_004 (RVRIRLVKRLLK)
+confirmed by external predictor screening (wave 0.5b; 2026-06-29).
+
+**Best candidates added to final panel (wave 0.5b, external predictor confirmed):**
+
+| Candidate | Sequence | AMPScanner | HemoFinder | AntiCP | Broad novelty (72-AMP DB) | Full-DB novelty |
+|-----------|----------|:----------:|:----------:|:------:|:-------------:|:---:|
+| SEED-020_VAR_004 | RLRIRVLKRLLK | 0.9928 (AMP) | LOW | Non-AntiCP 0.29 | NOVEL (41.7% max) | NOVEL |
+| SEED-020_VAR_002 | KVRIRVLKRLLK | 0.9960 (AMP) | LOW | Non-AntiCP 0.32 | NOVEL (37.5% max) | NOVEL |
+| SEED-019_VAR_004 | RVRIRLVKRLLK | panel anchor | LOW | Non-AntiCP | 38.1% max vs non-patent AMPs | **CLOSE_RELATIVE ⚠ patent risk** |
+
+> **IP NOTE — SEED-019_VAR_004:** Full 27,234-sequence novelty audit v2 (APD6, DRAMP general,
+> DRAMP patent, UniProt) classifies SEED-019_VAR_004 as **CLOSE_RELATIVE (66.7% identity to
+> DRAMP patent sequence DRAMP05502)**. **IP clearance is required before ordering synthesis
+> or publishing results for SEED-019_VAR_004.** The SEED-020 pair (VAR_004 and VAR_002) were
+> not flagged in the patent database and carry no identified IP risk.
+
+SEED-020_VAR_004 is the **best safety-activity tradeoff in the entire wave 0.5b batch** by
+combined external predictor consensus. SEED-020_VAR_002 has the highest AMPScanner score (0.9960)
+of any candidate in the batch.
+
+**Novelty:** SEED-020_VAR_004 and SEED-020_VAR_002 confirmed NOVEL by broad novelty check
+(72-AMP curated reference database; < 50% similarity). Closest matches are in the REF-RLK-001
+(cationic helix) and REF-LL37-003 (LL-37 fragment) classes — confirmed activity in either
+would be publishable. SEED-019_VAR_004 carries a patent proximity flag (see IP NOTE above);
+publishability claims for SEED-019_VAR_004 require IP clearance.
+
+**Mechanism:** Cationic amphipathic helix — alternating hydrophobic (Leu/Val/Ile) and cationic
+(Arg/Lys) faces interact with anionic bacterial membranes, causing rapid membrane depolarization.
+Expected to be Gram-negative and Gram-positive active (broad-spectrum). Each sequence carries
+3–4 Arg plus 2–3 Lys, for a combined Lys/Arg density of 6 in 12 AA; net charge at pH 7.4
+is approximately **+5 to +6** depending on terminal modifications — well above the +3 threshold
+for E. coli outer membrane permeation.
+
+**Hemolysis warning:** Macrel hemolysis predictor scores 0.93 for all SEED-020 variants.
+However, **Macrel hemolysis is non-discriminating in this dataset — 60/60 wave 0.5 candidates
+were flagged regardless of sequence class** (documented in evidence JSONs; see
+`docs/METRICS_CURRENT.md`). HemoFinder (a separate predictor) classifies all SEED-019/020
+variants as LOW hemolysis risk.
+**Run empirical HC50 assay — do not exclude on Macrel score alone.**
+
+**Synthesis guidance:**
+- No Met, no Cys → standard Fmoc SPPS, no oxidation risk, no disulfide risk.
+- **N-terminal acetylation (Nα-Ac) REQUIRED**: blocks N-terminal aminopeptidase (exopeptidase)
+  from degrading the free α-amine terminus (specify "N-terminal acetylation (Ac-)" in synthesis
+  order — zero extra cost). Note: Nα-Ac does NOT protect against endopeptidase (trypsin) cleavage
+  at the 5 interior K/R sites; endoproteolytic protection requires D-amino substitution (Wave 2).
+- **C-terminal amidation (CONH₂) RECOMMENDED**: adds +0.5 net charge, improves serum exopeptidase
+  resistance, standard for short cationic helices (specify "C-terminal amide" in synthesis order).
+- SHORT_PEPTIDE (12 AA): serum stability model calibrated for 18–30 AA — model score likely
+  **underestimates** actual t½; measure empirically (50% human serum, 37°C, 0/30/60/120 min HPLC).
+- 5 trypsin sites at R/K positions → WAVE2_D_AMINO substitution plan generated; if serum t½ < 30 min,
+  proceed directly to D-Arg substitution at positions R1, R3, R5 in Wave 2.
+- Storage: lyophilized at −20°C (no oxidation-sensitive residues; standard cationic peptide storage).
+
+**Assay priority:**
+- Screen SEED-020_VAR_004 and SEED-020_VAR_002 in the same plates as SEED-019_VAR_004 for
+  direct SAR comparison.
+- If both SEED-020 variants and SEED-019_VAR_004 are active, this SAR triad enables immediate
+  QSAR analysis of R→K and L→R substitution effects — a publishable structure-activity insight
+  even from a single wet-lab run.
+- MRSA USA300 is a priority target for this scaffold class (Arg-rich AMPs disrupt Gram-positive
+  membranes; no intracellular target required unlike SEED-009).
+
+---
+
 ## Serum Stability Model Limitations (All 7 Seeds)
 
 The `serum_stability_score()` function is calibrated for medium-length cationic helices (~18–30 AA).