Skip to content

Commit 9728339

Browse files
cschanhniemcschanhniem
authored
feat: Phase 4 — Boman scorer, consensus report, 89 AMP nominees, nomination report
Cherry-picked Phase 4 changes onto clean main: Boman index second scorer, scorer consensus batch pack section, 89 AMP nominees with evidence certificates, NOMINATION_REPORT.md scientific methodology, EXPERT_REVIEW_PACK.md live data update.
1 parent 75b17c8 commit 9728339

16 files changed

Lines changed: 1909 additions & 3 deletions

‎Makefile‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -50,7 +50,7 @@ generate:
5050
phase3: generate
5151
PYTHONPATH=src $(PYTHON) -m openamp_foundry.cli rank \
5252
--candidates examples/sequences/phase3_pool.csv \
53-
--references examples/sequences/amp_seeds.csv \
53+
--references examples/known_reference/amp_curated_references.csv \
5454
--out outputs/phase3_ranked.jsonl \
5555
--report outputs/phase3_report.md \
5656
--cert-dir outputs/phase3_evidence \

‎docs/EXPERT_REVIEW_PACK.md‎

Lines changed: 257 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,257 @@
1+
# OpenAMP Foundry — Expert Review Pack
2+
3+
**Phase 3 Candidate Batch**
4+
**Pipeline version:** 0.1.0 (v0.2 scorer update)
5+
**Generated:** 2026-06-27
6+
**Batch ID:** See `outputs/phase3_manifest.json`
7+
8+
---
9+
10+
## Purpose of This Document
11+
12+
This document is prepared for review by a qualified microbiology, peptide chemistry, or
13+
infectious disease expert prior to any wet-lab expenditure. It summarises the computational
14+
candidate nomination process, the selection rationale, and the recommended next steps.
15+
16+
**The expert reviewer is asked to:**
17+
1. Evaluate whether the candidate selection methodology is scientifically credible.
18+
2. Identify any obvious problems the computational pipeline may have missed.
19+
3. Advise on which candidates (if any) are worth synthesising.
20+
4. Recommend appropriate assay types and target organisms.
21+
5. Flag any safety or dual-use concerns.
22+
23+
---
24+
25+
## Computational Disclaimer
26+
27+
All scores in this document are heuristic physicochemical proxies. They are NOT validated
28+
biological predictors. No antimicrobial activity has been demonstrated in vitro or in vivo.
29+
Do not describe any candidate as an antibiotic, drug, cure, therapy, or proven antimicrobial
30+
without experimental evidence. The lab is the judge.
31+
32+
---
33+
34+
## 1. Overview of the Pipeline
35+
36+
The OpenAMP Foundry pipeline applies a six-stage scoring system to candidate sequences:
37+
38+
| Stage | Method | What it measures |
39+
|-------|--------|-----------------|
40+
| Activity-likeness | Physicochemical heuristics | Charge, hydrophobicity, amphipathicity, length |
41+
| Boman activity | Boman (2003) index normalized | Per-residue interaction potential; complements activity heuristic |
42+
| Model disagreement | \|activity − boman_activity\| | Scorer consensus: low = robust, high = uncertain |
43+
| Safety (toxicity proxy) | Physicochemical risk flags | Hemolysis-correlated excess hydrophobicity, charge density, repeats |
44+
| Synthesis feasibility | Composition and length filter | Cysteine content, proline content, repeat runs, length |
45+
| Novelty | Levenshtein distance | Sequence divergence from 45 known AMP references |
46+
| Ensemble | Weighted sum | activity×0.35 + safety×0.30 + synthesis×0.20 + novelty×0.15 |
47+
48+
**Pipeline configuration:** `configs/phase3.yaml`
49+
**Selection rule:** `docs/SELECTION_RULE.md` (locked before generation)
50+
**Full benchmark methodology:** `docs/BENCHMARKING.md`
51+
52+
**Key evidence level:** These are Level 0–2 evidence scores (syntax validity, reproducible features,
53+
transparent heuristics). They have not been validated against lab measurements. Lab assay
54+
(Level 5) is the required next gate.
55+
56+
---
57+
58+
## 2. Reference Set
59+
60+
Novelty was scored against **45 known antimicrobial peptides** drawn from published literature
61+
(see `examples/known_reference/amp_curated_references.csv`). Families represented:
62+
63+
- Magainin and pexiganan analogs (Zasloff 1987, Ge 1999)
64+
- Buforin family (Kim 1996, Park 2000)
65+
- Indolicidin and analogs (Selsted 1992)
66+
- Temporin family (Mangoni 2001, Simmaco 1996)
67+
- Aurein family (Rozek 2000)
68+
- Cecropin family (Steiner 1981, Hultmark 1980)
69+
- Cathelicidin-related short sequences
70+
- Melittin (as hemolysis positive control reference)
71+
72+
**Assessment:** Candidates with novelty < 0.10 against this reference set are near-duplicates
73+
of known AMPs and should be deprioritised. Novelty ≥ 0.30 indicates meaningful divergence from
74+
the known AMP landscape.
75+
76+
---
77+
78+
## 3. Generation Method
79+
80+
Candidates were produced by conservative amino-acid substitutions on 5 template seeds:
81+
82+
| Seed | Sequence | Family |
83+
|------|----------|--------|
84+
| SEED-001 | KWKLFKKIGAVLKVL | Magainin-like |
85+
| SEED-002 | GIGKFLHSAKKFGKAFVGEIMNS | Cecropin/Magainin-like |
86+
| SEED-003 | RRWQWRMKKLG | Cationic tryptophan-rich |
87+
| SEED-004 | FLPLIGRVLSGIL | Temporin-like |
88+
| SEED-005 | KRLFKKIGSALKFL | Hybrid cationic-hydrophobic |
89+
90+
Substitutions were restricted to physicochemically conservative groups (cationic→cationic,
91+
hydrophobic→hydrophobic, etc.) to preserve the amphipathic balance of the templates.
92+
93+
**Reviewer note:** This is an intentionally conservative generation strategy. It explores the
94+
near-neighbourhood of known AMP families. It does NOT explore truly novel sequence space.
95+
Future cycles should incorporate independent sequence generation (e.g. protein language models
96+
or Bayesian optimization) to achieve more radical novelty.
97+
98+
---
99+
100+
## 4. Batch Statistics
101+
102+
| Metric | Value |
103+
|--------|-------|
104+
| Sequences generated | 383 |
105+
| Sequences passing all filters | 89 |
106+
| Sequence length range | 11–14 aa |
107+
| Mean ensemble score | 0.781 |
108+
| Mean predicted activity | 0.839 |
109+
| Mean predicted safety | 1.000 |
110+
| Mean synthesis feasibility | 1.000 |
111+
| Mean novelty vs reference set | 0.172 |
112+
| Diversity clusters (threshold 0.80) | 32 |
113+
| Singleton clusters | 13 (40.6%) |
114+
| Mean Boman activity score | 0.503 |
115+
| Mean scorer disagreement | 0.311 |
116+
| High consensus (disagreement <0.20) | 1 |
117+
| Uncertain (disagreement ≥0.30) | 48 |
118+
119+
**Reviewer alert — Safety scores:** Mean safety = 1.0 because ALL selected candidates passed
120+
the safety filter (max_safety_risk = 0.40). This is because the conservative substitution
121+
generator preserves balanced charge/hydrophobicity. This looks good computationally, but does
122+
not replace wet-lab hemolysis and cytotoxicity assays.
123+
124+
**Reviewer alert — Scorer disagreement:** The mean disagreement between the activity-likeness
125+
heuristic and the Boman index second scorer is 0.311. This is expected: the activity scorer
126+
rewards amphipathic charge/hydrophobicity balance (good for membrane disruption), while the
127+
Boman index penalizes hydrophobic residues as lower interaction potential. The disagreement is
128+
scientifically honest — these candidates are predicted AMPs through amphipathic mechanisms but
129+
have moderate protein-binding potential by the Boman criterion. Candidates with higher Boman
130+
scores should be given extra weight in expert review.
131+
132+
**Reviewer alert — Cluster concentration:** 59% of candidates fall within multi-member clusters,
133+
meaning significant chemical similarity within the batch. Only ~13 candidates are sequence-isolated
134+
singletons. The expert should advise whether this diversity level is adequate for meaningful assay.
135+
136+
---
137+
138+
## 5. Top 20 Candidates by Ensemble Score (with Dual-Scorer Columns)
139+
140+
> These candidates passed all computational filters and were selected by greedy diversity.
141+
> Rankings are provisional and may change with improved scoring in future pipeline versions.
142+
> **Dis** = scorer disagreement; lower is more robust (dual-scorer consensus).
143+
144+
| Rank | Candidate ID | Sequence | Ens | Activity | Boman | Dis | Safety | Novelty | Len |
145+
|------|-------------|----------|-----|----------|-------|-----|--------|---------|-----|
146+
| 1 | SEED-005_VAR_049 | KRLFKKIPSALKFF | 0.880 | 0.885 | 0.485 | 0.400 | 1.000 | 0.467 | 14 |
147+
| 2 | SEED-005_VAR_009 | KRFFKKIGSALKFA | 0.877 | 0.878 | 0.508 | 0.370 | 1.000 | 0.467 | 14 |
148+
| 3 | SEED-005_VAR_063 | KRLFRKIGSALKFV | 0.874 | 0.867 | 0.503 | 0.365 | 1.000 | 0.467 | 14 |
149+
| 4 | SEED-005_VAR_058 | KRLFKKVGSALRFL | 0.871 | 0.861 | 0.503 | 0.358 | 1.000 | 0.467 | 14 |
150+
| 5 | SEED-005_VAR_023 | KRLFKKIGRALKFL | 0.870 | 0.913 | 0.537 | 0.376 | 1.000 | 0.333 | 14 |
151+
| 6 | SEED-005_VAR_061 | KRLFKRLGSALKFL | 0.868 | 0.853 | 0.497 | 0.355 | 1.000 | 0.467 | 14 |
152+
| 7 | SEED-005_VAR_068 | KRLMKKIGSAIKFL | 0.856 | 0.818 | 0.519 | 0.299 | 1.000 | 0.467 | 14 |
153+
| 8 | SEED-005_VAR_013 | KRLAKHIGSALKFL | 0.855 | 0.814 | 0.480 | 0.334 | 1.000 | 0.467 | 14 |
154+
| 9 | SEED-005_VAR_002 | HRLIKKIGSALKFL | 0.849 | 0.813 | 0.457 | 0.356 | 1.000 | 0.429 | 14 |
155+
| 10 | SEED-003_VAR_027 | RRWQWRFKRLG | 0.839 | 0.890 | 0.532 | 0.358 | 1.000 | 0.182 | 11 |
156+
| 11 | SEED-003_VAR_045 | RRWQYRIKKLG | 0.834 | 0.877 | 0.572 | 0.305 | 1.000 | 0.182 | 11 |
157+
| 12 | SEED-003_VAR_020 | RRWQWHIKKLG | 0.832 | 0.870 | 0.480 | 0.390 | 1.000 | 0.182 | 11 |
158+
| 13 | SEED-003_VAR_014 | RRFQWRMRKLG | 0.831 | 0.867 | 0.579 | 0.288 | 1.000 | 0.182 | 11 |
159+
| 14 | SEED-003_VAR_017 | RRWNWRMRKLG | 0.829 | 0.862 | 0.559 | 0.303 | 1.000 | 0.182 | 11 |
160+
| 15 | SEED-005_VAR_047 | KRLFKKIGSVLKML | 0.829 | 0.825 | 0.501 | 0.324 | 1.000 | 0.267 | 14 |
161+
| 16 | SEED-003_VAR_048 | RRWSWHMKKLG | 0.828 | 0.860 | 0.493 | 0.367 | 1.000 | 0.182 | 11 |
162+
| 17 | SEED-003_VAR_003 | KRWQWRIKKLG | 0.828 | 0.859 | 0.547 | 0.311 | 1.000 | 0.182 | 11 |
163+
| 18 | SEED-003_VAR_049 | RRWSWRMHKLG | 0.828 | 0.859 | 0.493 | 0.365 | 1.000 | 0.182 | 11 |
164+
| 19 | SEED-003_VAR_051 | RRWTWRMKKAG | 0.828 | 0.858 | 0.592 | 0.266 | 1.000 | 0.182 | 11 |
165+
| 20 | SEED-003_VAR_042 | RRWQWRMRKLP | 0.828 | 0.858 | 0.559 | 0.299 | 1.000 | 0.182 | 11 |
166+
167+
Full list: `outputs/phase3_report.md`
168+
Machine-readable details: `outputs/phase3_batch_pack.json`
169+
Evidence certificates: `outputs/phase3_evidence/` (89 selected, all schema-validated)
170+
171+
**Reviewer priority note:** The SEED-003 family (tryptophan-rich 11-mers) has higher Boman
172+
activity scores and lower disagreement than SEED-005 (14-mers), suggesting it should be prioritised
173+
for synthesis. Rank 19 (RRWTWRMKKAG) has the highest Boman score among the top-20 (0.592) with
174+
only moderate disagreement (0.266).
175+
176+
---
177+
178+
## 6. Reviewer Questions
179+
180+
Please advise on the following:
181+
182+
### 6.1 Scientific credibility
183+
- [ ] Is the physicochemical scoring approach defensible for pre-screening short AMPs?
184+
- [ ] Is the novelty threshold (≥ 0.05 against 45 known AMP references) meaningful?
185+
- [ ] Does the diversity selection (max pairwise similarity 0.80) provide adequate batch diversity?
186+
- [ ] Is the Boman index (2003) a defensible second scorer for this class of peptides?
187+
- [ ] Is the mean scorer disagreement (0.311) consistent with what you would expect for helical AMPs?
188+
189+
### 6.2 Candidate quality
190+
- [ ] Do any of the top-20 candidates have obvious structural or sequence-level problems?
191+
- [ ] Are there specific sequences in the batch you would prioritise or deprioritise?
192+
- [ ] Is 11–14 aa the right length range, or should we explore longer candidates?
193+
- [ ] Do the SEED-003 tryptophan-rich 11-mers look more promising than the SEED-005 14-mers?
194+
195+
### 6.3 Assay recommendations
196+
- [ ] Which assay types are most appropriate for initial screening? (Suggested: MIC, hemolysis)
197+
- [ ] Which bacterial strains/organisms should be tested against?
198+
- [ ] What constitutes a meaningful positive result? (e.g. MIC ≤ 8 µg/mL)
199+
- [ ] How many candidates can reasonably be tested in a first-round assay?
200+
- [ ] Do you have a preferred CRO or academic partner for peptide assays?
201+
202+
### 6.4 Safety and dual-use
203+
- [ ] Are any sequences potentially concerning from a biosafety perspective?
204+
- [ ] Are there any concerns about these sequences being misused?
205+
206+
---
207+
208+
## 7. Recommended Next Steps
209+
210+
Before any candidate proceeds to synthesis or assay:
211+
212+
1. **Expert sign-off**: A qualified reviewer completes section 6 above.
213+
2. **Synthesis plan**: A peptide chemist reviews feasibility and recommends synthesis vendor.
214+
3. **Target organism selection**: Microbiologist selects appropriate test organisms.
215+
4. **Protocol pre-registration**: Assay protocol and pass/fail criteria locked before synthesis.
216+
5. **Negative controls**: Known inactive sequences (e.g. scrambled versions) included in assay.
217+
6. **Ethical and regulatory check**: PI confirms no IRB, ITAR, dual-use, or export-control issues.
218+
7. **Pilot assay**: Small pilot (top 5–10 candidates by Boman+activity consensus) before full batch.
219+
220+
**Suggested pilot candidates (highest Boman × lowest disagreement):**
221+
222+
| Priority | Candidate ID | Sequence | Boman | Dis | Rationale |
223+
|----------|-------------|----------|-------|-----|-----------|
224+
| 1st | SEED-003_VAR_051 | RRWTWRMKKAG | 0.592 | 0.266 | Highest Boman among top-20, low disagreement |
225+
| 2nd | SEED-003_VAR_014 | RRFQWRMRKLG | 0.579 | 0.288 | Strong Boman, low disagreement |
226+
| 3rd | SEED-003_VAR_045 | RRWQYRIKKLG | 0.572 | 0.305 | Good Boman, 11-mer |
227+
| 4th | SEED-003_VAR_003 | KRWQWRIKKLG | 0.547 | 0.311 | Good Boman, varied flanking residue |
228+
| 5th | SEED-005_VAR_023 | KRLFKKIGRALKFL | 0.537 | 0.376 | Highest activity in batch (0.913) |
229+
230+
---
231+
232+
## 8. Pipeline Limitations (Required Disclosure)
233+
234+
The expert reviewer must be aware of these limitations:
235+
236+
| Limitation | Impact |
237+
|-----------|--------|
238+
| Level 0–2 evidence only | Scores are physicochemical heuristics, not validated ML predictions |
239+
| Two scorers only | Boman index added as second scorer; disagreement available; no ML predictor yet |
240+
| High mean disagreement (0.311) | Scorers model AMP activity by different mechanisms; consensus is weak for this batch |
241+
| Novelty vs. 45 references only | Novel against this reference set does not mean novel against all known AMPs |
242+
| Conservative generation only | Near-seed variants; genuinely novel families not explored |
243+
| No structural prediction | No secondary structure, membrane interaction, or 3D modeling |
244+
| Amphipathicity is helical-only | Assumes helix; β-sheet or other motifs not modeled |
245+
| Safety proxy is not validated | The pipeline has never been calibrated against hemolysis data |
246+
| Selection is single-objective | No multi-objective Pareto optimization; ensemble weights are heuristic |
247+
248+
---
249+
250+
## Contact
251+
252+
This document is produced by the OpenAMP Foundry computational pipeline.
253+
For human contact, see the repository README for contributor information.
254+
All decisions about synthesis, assay, or publication require human expert authorization.
255+
256+
**No candidate may be synthesised or sent to any external party without qualified expert review
257+
and explicit human approval.**

0 commit comments

Comments
 (0)