cschanhniem · Jun 28, 2026 · Jun 28, 2026 · Jun 28, 2026
diff --git a/docs/ROADMAP.md b/docs/ROADMAP.md
@@ -81,7 +81,7 @@ review (2026-06-28). Progress on these would materially raise breakthrough proba
 | Gap | Why it matters | Effort estimate |
 |-----|----------------|-----------------|
 | Large-scale benchmark (≥ 500 AMPs vs composition-matched decoys, cluster-split) | Current AUROC 0.8420 measured on 43+44 demo set (n=87, CI₉₅: 0.76–0.91); may not generalise | Medium |
-| External predictor ensemble adapters (AMPScanner, AntiCP2, AMPlify, Macrel) | Independent second opinions on activity; required for scientific credibility | Medium |
+| External predictor ensemble adapters (CAMPR4, AMPScanner, dbAMP, AntiCP2, Macrel) | Independent second opinions on activity; required for scientific credibility. Manual web-submission checklist at `outputs/external_predict_checklist.md`. Macrel v1.6.0 CLI ONNX bug documented PR #77 — all sequences (incl. canonical AMPs magainin-2, LL-37) misclassified as NAMP; use web server at big-data-biology.org/software/macrel. AMPlify omitted: GPU/ONNX env incompatible with current deps | Medium (checklist generated; web submissions pending) |
 | True novelty check against APD3, DRAMP v3.0, dbAMP, UniProt | Current novelty scored against 45-sequence seed set only; may overestimate novelty | Small–Medium |
 | ~~AUPRC alongside AUROC~~ | ~~Better metric for class-imbalanced AMP datasets~~ | **Done** (PR #58; updated PR #72) — pipeline AUPRC = 0.8627 |
 | Wet-lab result integration (active-learning round 2) | Required to move from 15–30% to 50%+ credible probability | Requires wet-lab |

diff --git a/docs/WET_LAB_HANDOFF.md b/docs/WET_LAB_HANDOFF.md
@@ -132,7 +132,7 @@ Hemolysis > 10% at MIC is a fail.
 
 SEED-006 variants are derived from **Mastoparan-X** (INWKGIAAMAKKLL, Yoshida *et al.* 1990),
 a wasp (*Vespa xanthoptera*) venom calmodulin-binding helical peptide. This is structurally
-distinct from the other 5 seeds (magainin-like, cecropin-like, tachyplesin-like). 
+distinct from the other 6 seeds (magainin-like, cecropin-like, tachyplesin-like, Bac2A proline-rich, puroindoline Trp-rich). 
 
 **Mechanism:** Mastoparan-X inserts into bacterial membranes via a predominantly
 **hydrophobic helix insertion** mechanism rather than electrostatic carpet disruption.
@@ -156,17 +156,71 @@ It also activates G-proteins and has calmodulin-binding properties.
 
 ---
 
-## Serum Stability Model Limitations (SEED-003, SEED-005, SEED-008)
+## SEED-001 (Magainin-1) Special Notes
+
+SEED-001 variants are derived from **Magainin-1** (GIGKFLHSAKKFGKAFVGEIMKS, Zasloff 1987, PNAS),
+a 23-AA cationic α-helical AMP from *Xenopus laevis* (African clawed frog) skin. Our variant
+SEED-001_VAR_064 (KWKLFRKIGAVLRVL, 15 AA) is a truncated, charge-optimised N-terminal segment
+re-entered into the synthesis pool as the 7th scaffold family in PR #72 (face_segregation_bonus).
+
+**Mechanism:** Toroidal-pore (wormhole) mechanism — the primary mode supported by solid-state NMR:
+cationic helix is electrostatically recruited to the bacterial membrane (negative surface charge),
+reorients perpendicular to the bilayer, and forms transient pores in which lipid headgroups line
+the channel alongside the peptide (Matsuzaki et al. 1996, Biochemistry 35:11361; Huang 2006,
+Biophys J 90:L6). At high concentrations or in truncated variants, a carpet-like detergent
+mechanism may also operate (Brogden 2005, Nat Rev Microbiol 3:238). The two mechanisms differ in
+assay signature: toroidal pores produce graded voltage-dependent conductance; carpet mechanism
+produces a threshold-like all-or-nothing lysis. Observe dose-response shape to distinguish.
+
+**Novelty:** LOW (0.133 vs reference set). Magainin-2 is in the reference set.
+SEED-001_VAR_064 is a known-family derivative, not a novel scaffold.
+**Its primary value is as a positive control confirming helix-mechanism activity at 15 AA length.**
+
+**Scorer disagreement:** SEED-001_VAR_064 has the highest disagreement score in the pilot panel:
+activity_likeness = 0.8056 (physicochemical scorer), boman_activity = 0.4286 (Boman index),
+disagreement = 0.377 (gate threshold 0.45). The Boman index classifies this peptide as borderline
+while the physicochemical scorer classifies it as strongly active. If SEED-001_VAR_064 is inactive
+in assay, this internal disagreement is the predicted explanation — report it alongside the result.
+
+**Expected activity profile:**
+- Gram-negative and Gram-positive coverage expected (toroidal-pore/carpet disruption is broad-spectrum)
+- Dose-response typically sigmoidal with a clear MIC breakpoint
+- Hemolytic risk: moderate-high at high concentrations; MIC/3 starting concentration (μH = 0.53)
+
+**Synthesis guidance:**
+- No Cys, no Met → standard Fmoc SPPS, standard storage conditions
+- **N-terminal acetylation strongly recommended** (5 K/R sites including K1 at N-terminus;
+  Ac- blocks aminopeptidase entry and improves serum stability)
+- **C-terminal amidation recommended** (`CONH₂`; adds +0.7 charge, improves serum stability)
+- Serum stability model score 0.49 (borderline) — see stability section below; natural magainin
+  t½ ~15–30 min in serum; N-Ac/C-NH₂ modifications are known to reduce exopeptidase exposure
+  and may meaningfully extend t½ (assay with and without modifications to quantify)
+
+**Assay interpretation:**
+- If SEED-001_VAR_064 is inactive at ≤ 128 µg/mL: check HPLC purity (aggregation risk at
+  shorter length); retest with 0.01% BSA to prevent tube adhesion
+- If SEED-001_VAR_064 is active but MIC > 8 µg/mL: expected for a 15-mer vs. native 22-mer;
+  Wave 2 should test longer variants (17–20 AA) of the same sequence
+- **Target MIC ≤ 10 µg/mL**: below magainin-2 reference MIC (the 23-mer benchmark); if
+  achieved, the truncation to 15 AA is biologically significant and worth reporting
+- If all 6 novel seeds are inactive but SEED-001 is active: confirms the pipeline scoring
+  works mechanistically, even if the novel scaffolds failed — still a useful calibration outcome
+
+---
+
+## Serum Stability Model Limitations (SEED-001, SEED-003, SEED-005, SEED-008, SEED-009)
 
 The `serum_stability_score()` function is calibrated for medium-length cationic helices (~18–30 AA).
-All 6 pilot panel seed families have model scores below the 0.70 threshold (t½ > 2h gate).
-Three families are flagged as borderline or likely model underestimates:
+All 7 pilot panel seed families have model scores below the 0.70 threshold (t½ > 2h gate).
+Five families are flagged as borderline or carrying a known model bias:
 
 | Scaffold | Pilot score | Issue | Expected actual t½ | Action |
 |----------|-------------|-------|-------------------|--------|
-| SEED-003 (cationic helix, 11–14 AA) | 0.35–0.38 | Short peptide has fewer cleavage sites than the model's per-length baseline — systematic underestimate | **Likely > model prediction** — Radzishevsky et al. (2007, Nat Biotechnol): short AMPs show 2–5× longer t½ than length-matched controls predict | Do not exclude on serum score alone. Run assay and compare to model |
+| SEED-003 (cationic helix, 11–14 AA) | 0.35–0.38 | Short peptide: the trypsin-site density model was calibrated on 18–30 AA sequences. For <15 AA peptides, per-site cleavage probability and flanking-context effects dominate over the aggregate density metric; model scores below 0.40 are unreliable at this length | **Likely > model prediction** (known calibration range limitation — not a specific literature citation) | Do not exclude on serum score alone. Run assay and compare to model |
 | SEED-005 (cecropin-magainin hybrid, 14 AA) | 0.449 | Borderline score below 0.50 (t½ > 1h) threshold; only 1 pilot slot — any failure drops this family | ~30 min–1 h by model; no documented correction factor | Prioritise in early serum screen; D-Lys Wave 2 plan ready if needed |
-| SEED-008 (puroindoline-a, Trp-rich) | 0.45 | Indole ring bulk reduces chymotrypsin cleavage at Trp4 (steric interference) | **Likely > model prediction** — Wu & Ding (2016, J Pept Sci): Trp-flanked cleavage sites are cut 3–8× slower than Tyr/Phe equivalents | D-Trp Wave 2 plan stands; assay actual stability first — may not need D-modification |
+| SEED-008 (puroindoline-a, Trp-rich) | 0.43–0.47 | Indole ring bulk reduces chymotrypsin cleavage at Trp4 (steric interference) | **Likely > model prediction** — Wu & Ding (2016, J Pept Sci): Trp-flanked cleavage sites are cut 3–8× slower than Tyr/Phe equivalents | D-Trp Wave 2 plan stands; assay actual stability first — may not need D-modification |
+| SEED-001 (magainin-1 derivative, 15 AA) | 0.49 | 5 K/R sites in 15-mer; natural magainin-2 has documented serum t½ 15–30 min (Park & Hahm 2005, Curr Protein Pept Sci) — score may be accurate, not a model underestimate | ~30–60 min estimated | N-terminal acetylation (recommended in QC) partially protects K1; run early serum screen before committing MIC budget; D-Lys at K3/K7 in Wave 2 if t½ < 30 min |
+| SEED-009 (Bac2A proline-rich, 12 AA) | 0.57 | Pro residues confer resistance to trypsin, chymotrypsin, and elastase: Pro-X bonds cannot be hydrolysed (Pro nitrogen lacks a free H for the serine protease oxyanion hole mechanism); X-Pro bonds are very slowly cleaved due to cis-trans isomerisation barriers (Vanhoof et al. 1995, FASEB J 9:736–744). The trypsin-site density model counts K/R residues only and cannot capture this resistance. | **Likely > model prediction** — Bac2A-type Pro-rich AMPs show > 2h stability in 50% human serum (Otvos & Cudic 2002, Curr Pharm Design) | Do not penalise SEED-009 for low model score; assay confirms likely robustness |
 
 **Recommended action before full MIC panel:**
 Run a serum stability screen on all 20 pilot candidates:
@@ -175,8 +229,8 @@ Run a serum stability screen on all 20 pilot candidates:
 - Cost: ~$200–400 per batch of multiple candidates (not per individual peptide)
 
 This validates the gate assumptions directly and prevents retiring candidates that the model
-incorrectly penalises. For SEED-006 and SEED-007 (scores 0.61–0.67), the assay will
-determine whether borderline scores translate to the required t½ > 2h.
+incorrectly penalises. For SEED-006 (scores 0.61–0.67) and SEED-007 (scores 0.64–0.66), the
+assay will determine whether borderline scores translate to the required t½ > 2h.
 
 ---
 
@@ -196,7 +250,7 @@ hydrophilic faces and reports face contrast, cationic segregation, and amphipath
 | SEED-006 (mastoparan-X, 14 AA) | 0.960 | 0.480 | 0% | Helical insertion | ✅ Good amphipathic architecture |
 | SEED-007 (bombolitin II, 18 AA) | 0.978 | 0.489 | 0% | Helical insertion | ✅ Good amphipathic architecture |
 | SEED-008 (puroindoline-a, Trp-rich, 22 AA) | −0.139 | 0.000 | 8% | **Aromatic anchoring** | ⚠️ Not a classical helix-wheel AMP — expected |
-| SEED-009 (Bac2A proline-rich, 9 AA) | 0.036 | 0.018 | 50% | **Intracellular (ribosome)** | ⚠️ Not a classical helix-wheel AMP — expected |
+| SEED-009 (Bac2A proline-rich, 12 AA) | 0.036 | 0.018 | 50% | **Intracellular (ribosome)** | ⚠️ Not a classical helix-wheel AMP — expected |
 
 > **Note:** Amphipathic scores depend on the N-terminal convention (position-0 residue). All pipeline
 > candidates are presented N→C as synthesized. Do not compare scores across circularly permuted