Version: 0.2.x (pipeline version); document updated for PRs #47–78
Status: Working draft for expert review. Not a peer-reviewed publication.
We describe a transparent, reproducible dry-lab pipeline for computational nomination of short antimicrobial peptide (AMP) candidates. The pipeline applies physicochemical heuristic scoring, novelty analysis against a curated reference set, synthesis feasibility filtering, and diversity selection to produce a batch of computationally nominated candidates with machine-readable evidence certificates. All scores are heuristic proxies; none have been calibrated against experimental measurements. Candidates are nominated for possible expert review and qualified wet-lab assay. No biological activity claims are made.
Candidates were produced by the template_mutator.py generator, which applies conservative
amino-acid substitutions to five AMP-like template sequences. Substitutions are restricted to
physicochemically conservative groups:
| Group | Amino acids |
|---|---|
| Cationic | K, R, H |
| Aliphatic/hydrophobic | L, I, V, A, F, M |
| Aromatic | F, W, Y |
| Polar/neutral | S, T, N, Q |
| Anionic | D, E |
| Conformational | G, P |
| Disulfide | C (no substitutions) |
Three variant strategies were applied:
- Single substitutions: All single-position conservative replacements (deterministic)
- Double substitutions: Random pairs of conservative positions (n=25 per seed, rng_seed=2024)
- Charge-enhanced variants: Polar positions (S, T, N, Q) replaced by K or R (n=12 per seed)
Total: 383 unique variants from 5 seeds (rng_seed=2024).
Note (PRs #63–110): The generation strategy above (5 seeds, rng_seed=2024) describes the initial variant corpus. Scoring was then improved across PRs #47–72. Benchmark expanded in PR #110 from 43+44 (n=87) to 95 AMPs + 96 decoys (n=191): the expanded benchmark achieves AUROC 0.7832 (bootstrap CI₉₅: 0.72–0.84). The original demo set (43 AMPs, n=87) achieved AUROC 0.8420 and remains the historical baseline. The expanded set includes 52 additional well-characterised public-domain AMPs from 12 taxonomic classes and provides a more honest generalisation estimate. Separately, the synthesis pool was expanded to 7 scaffold families: SEED-006 through SEED-009 added in expansion PRs #38–58; SEED-001 (magainin-1) re-entered PR #72. Current synthesis pool: 100 candidates from 7 scaffold families. Same pipeline config (
pipeline.yaml,phase3.yaml); wider seed set.
Limitation: This generation strategy produces near-seed variants. It does not explore genuinely novel sequence space. Future work should incorporate protein language model sampling or Bayesian sequence optimization.
Sequences were validated against the following criteria:
- Length 8–35 amino acids (inclusive)
- Canonical amino acid alphabet only (A, C, D, E, F, G, H, I, K, L, M, N, P, Q, R, S, T, V, W, Y)
Sequences failing either criterion received zero scores for activity and safety, and were excluded from the selected batch.
The following features were extracted for each sequence:
| Feature | Formula | Reference |
|---|---|---|
| Length | Count of residues | — |
| Net charge proxy | (K+R+H) − (D+E) | Wimley & White 1996 |
| Charge density | net_charge / length | — |
| Hydrophobic fraction | (L+I+V+A+F+M+W) / length | Kyte & Doolittle 1982 |
| Aromatic fraction | (F+W+Y) / length | — |
| Cysteine fraction | C / length | — |
| Glycine fraction | G / length | — |
| Proline fraction | P / length | — |
| Longest repeat run | Longest run of identical consecutive residues | — |
| Hydrophobic moment | Eisenberg (1984), 100°/residue helical angle | Eisenberg et al. 1984 |
| Boman index | Mean per-residue interaction potential | Boman 2003, Table 1, set 2 |
| GRAVY | Grand Average of hYdropathicity; mean Kyte-Doolittle value | Kyte & Doolittle 1982 |
Limitation: Features assume linear, unstructured sequence. Amphipathicity (hydrophobic moment) assumes an α-helical conformation (100°/residue). Non-helical AMPs are not modeled.
A heuristic score approximating AMP-like physicochemical properties. Computed from
eight terms; weights encode biochemical priors only (not data-optimised). Pre-clamp ceiling
1.02; final clamp01() bounds output to ≤ 1.0. Without helix_wheel_amphipathic_score key
the pre-clamp ceiling is 0.97 (backward-compatible with pre-PR-#72 feature dicts).
| Term | Contribution | Literature basis |
|---|---|---|
| Length score | × 0.24 | Peak at 18 aa; broad tolerance ±25 aa (Jenssen 2006) |
| Charge density (pH 7.4) | × 0.27 | +2 to +9 net charge for active AMPs (Zasloff 2002) |
| Hydrophobic fraction | × 0.17 | 40–50% sweet spot for membrane interaction (Hancock 2006) |
| Aromatic bonus | up to +0.10 | F/W/Y aid membrane insertion; capped at 20% fraction |
| Amphipathicity (μH) | up to +0.14 | μH > 0.4 correlates with membrane disruption (Eisenberg 1984) |
| Helix bonus (Pα) | up to +0.03 | Chou-Fasman Pα > 1.0; helical AMPs ~70% of membrane-active families (Huang 2000) |
| Cross-term (cd × μH) | up to +0.02 | Electrostatic + hydrophobic insertion synergy; normaliser 0.15 |
| Face segregation (hw_amp) | up to +0.05 | helix_wheel_amphipathic_score from moment-oriented wheel; [0,1] |
Formula (Python source: scoring/activity.py):
# Anionic guard (PR #70): if charge_density_ph74 < 0.0, return 0.0 immediately.
# Anionic peptides are electrostatically repelled by bacterial membranes.
μH_eff = max(hydrophobic_moment, max_hydrophobic_moment) # windowed mu_h (PR #70)
score = 0.24 × length_score
+ 0.27 × clamp01((charge_density_ph74 + 0.05) / 0.55)
+ 0.17 × (1 − min(|hydrophobic − 0.45| / 0.45, 1.0))
+ min(aromatic / 0.20, 1.0) × 0.10 # aromatic_bonus
+ clamp01(μH_eff / 0.8) × 0.14 # amphipathicity_score (windowed)
+ clamp01((Pα − 1.0) / 0.20) × 0.03 # helix_bonus
+ clamp01(charge_density_ph74 × μH_eff / 0.15) × 0.02 # cross_bonus
+ helix_wheel_amphipathic_score × 0.05 # face_segregation_bonus (PR #72)
# Pre-clamp ceiling: 0.24+0.27+0.17+0.10+0.14+0.03+0.02+0.05 = 1.02
# clamp01(score) at the end clips to [0, 1]. Without face_segregation key the ceiling is 0.97.
μH_eff = max(hydrophobic_moment, max_hydrophobic_moment). max_hydrophobic_moment is the
best μH over any 11-residue window (Eisenberg standard). For sequences ≤ 11 AA this equals
the full-sequence μH. For longer sequences it captures the most amphipathic helical segment,
which better reflects the membrane-active region. (Added PR #70.)
helix_wheel_amphipathic_score is the face-contrast score from helix_wheel_faces() (PR #71),
computed as max(0, min(1, face_contrast / 2.0)) where face_contrast is the mean Eisenberg
hydrophobicity difference between the hydrophobic and hydrophilic helical faces. This measures
whether hydrophobic and cationic residues are properly segregated onto opposite helical faces —
a complementary signal to μH (moment magnitude alone). Falls back to 0.0 when absent (backward-
compatible). (Added as scoring term PR #72; computed by helix_wheel_faces() in PR #71.)
charge_density_ph74 is the net peptide charge at pH 7.4 (Henderson-Hasselbalch) divided
by sequence length; the code falls back to charge_density (integer net-charge proxy / length)
for callers that pre-date the pH-7.4 charge model. Lookup uses explicit if key in features
conditional (not dict.get()) to avoid Python's eager evaluation of the fallback argument.
A second, independent activity score derived from the Boman index (Boman 2003):
Boman index = mean of per-residue interaction potentials
Interaction potentials are from Boman (2003), Table 1, set 2. Positive Boman index (> 1.0) correlates with AMP-like protein-binding potential in the published benchmark. The raw index is normalized to [0, 1] using a tanh mapping centred at 0.
Model disagreement score: disagreement = |activity_likeness − boman_activity|
A high disagreement value (≥ 0.30) indicates that the two independent scorers disagree on the candidate's activity potential. This is a computational uncertainty signal, not a biological risk flag. High-disagreement candidates warrant additional scrutiny before lab nomination. Low-disagreement candidates (both scorers agree) are computationally more robust nominations.
Reference: Boman HG (2003). Peptide antibiotics and their role in innate immunity. Annual Review of Immunology, 13, 61-92.
Coarse hemolysis/cytotoxicity risk proxy. Higher score = safer. Computed from six risk
signals that sum to a risk variable; score = 1.0 − clamp(risk, 0, 1).
| Risk signal | Threshold | Risk increment | Literature basis |
|---|---|---|---|
| Strong amphipathicity (μH) | μH > 0.55 | (μH − 0.55) × 1.5 | Non-selective membrane disruption (Dathe 1999 BBA) |
| Excess hydrophobicity | hydrophobic_fraction > 0.65 | (hf − 0.65) × 1.8 | Membrane partitioning at high hf |
| Excess charge density (pH 7.4) | charge_density_ph74 > 0.55 | (cd − 0.55) × 1.2 | Non-selective cationic disruption |
| Long sequence | length > 35 | +0.25 | Synthesis yield loss |
| High cysteine content | cysteine_fraction > 0.25 | +0.20 | Disulfide-mediated aggregation |
| Long repeat run | longest_repeat_run ≥ 6 | +0.15 | On-resin aggregation |
charge_density_ph74 (signed) is used directly — anionic sequences (charge_density_ph74 < 0)
receive a safety score of 1.0 (no hemolysis concern) because negatively charged peptides are
electrostatically repelled from anionic bacterial membranes. The abs() fallback was removed
in PR #65 to avoid incorrectly penalising anionic peptides as "hemolytic".
Critical limitation: This score has never been calibrated against experimental hemolysis or cytotoxicity measurements. It is a physics-based proxy only. Lab hemolysis assay (e.g., red blood cell lysis assay) is mandatory before any safety claim.
Penalizes sequences predicted to be difficult to synthesize by solid-phase peptide synthesis (SPPS). Score starts at 1.0; each penalty is subtractive; final score clamped to [0, 1].
| Penalty | Threshold | Deduction | Mechanism |
|---|---|---|---|
| Long sequence | length > 30 | min((length − 30) × 0.04, 0.40) | Deletion error accumulation; up to −0.40 at length ≥ 40 |
| Very short sequence | length < 8 | −0.30 flat | Unreliable purification / characterisation |
| Long homo-repeat run (any AA) | longest_repeat_run ≥ 5 | −0.10 flat | Coupling efficiency drops on identical-residue stretches |
| High cysteine fraction | cysteine_fraction > 0.20 | −0.15 flat | Disulphide scrambling; side-chain protection cost |
| Aggregation propensity | aggregation_propensity > 0 | min(prop × 0.25, 0.20) | On-resin self-association from hydrophobic runs {V,I,L,M,F,W} or β-branched density |
| High proline fraction | proline_fraction > 0.15 | −0.10 flat | N-terminal DKP; slow coupling at XP junctions during Fmoc deprotection |
Proline penalty detail: Proline's N-methylated backbone causes slow coupling at XP-junction positions and N-terminal diketopiperazine (DKP) cyclisation risk during piperidine Fmoc deprotection. References: Barlos et al. (1989) Int J Peptide Protein Res; Quibell et al. (1994) J Am Chem Soc; Fischer (2003) Curr Opin Drug Discov Devel.
Aggregation penalty detail: aggregation_propensity from features/physchem.py uses
two components: (1) longest consecutive run in {V,I,L,M,F,W} (ramp: 0 at run < 4, 1.0 at
run ≥ 8) and (2) β-branched density (V,I,T > 20%). The synthesis penalty is
min(agg × 0.25, 0.20) — a linear ramp from zero, capped at −0.20.
Length boundary note: The synthesis penalty threshold (length > 30) differs from the pipeline's sequence validity filter (length > 35 is flagged in safety, not synthesis). A candidate of length 32 incurs a −0.08 synthesis deduction but is not safety-penalised for length.
Estimates relative proteolytic longevity from interior cleavage-site densities. Higher = more
stable. Score = 1 − clamp(combined_density / 0.5, 0, 1).
| Protease | P1 substrates | Site weight | Literature |
|---|---|---|---|
| Trypsin | K, R (interior) | 2.0 | Dominant serum degradation; t½ < 30 min at ≥3 interior K/R (Hilpert 2006) |
| Chymotrypsin | F, W, Y (interior) | 1.0 | Secondary serum route |
| Neutrophil elastase (HNE) | A, V, S (interior) | 0.5 | Abundant at infection sites; HNE cleaves Ala > Val > Ser (Bieth 1986; Doherty 1991) |
combined_density = (2.0 × trypsin + 1.0 × chymo + 0.5 × elastase) / 3.5
Score 1.0 = no interior sites; 0.5 = combined density ≈ 0.25/residue; 0.0 ≥ 0.5/residue. This score does not gate the ensemble — it is used only in the pilot panel ranking (§5.2).
Novelty is computed as: 1 - max_similarity, where max_similarity is the maximum normalized
Levenshtein similarity between the candidate and any sequence in the 72-sequence reference set
(examples/known_reference/amp_curated_references.csv).
Normalized Levenshtein similarity: 1 - (edit_distance / max_length).
A novelty of 0.0 means the sequence is identical to a known reference. A novelty of 1.0 means no sequence in the reference set is similar at all.
Limitation: The reference set contains only 72 curated sequences. This is not a complete survey of known AMPs. Real-world novelty should be checked against APD3, CAMP, DRAMP, and UniProt antimicrobial sequences.
ensemble = 0.35 × activity + 0.30 × safety + 0.20 × synthesis + 0.15 × novelty
Weights were set by expert judgment before any candidate generation. They are not
data-optimized and have not been validated. The scoring rule is pre-registered in
docs/evidence/SELECTION_RULE.md.
- Filter: Length 8–35 aa, canonical AAs only, safety ≥ 0.60, novelty ≥ 0.05, ensemble ≥ 0.0
- Rank: By ensemble score (descending)
- Diversity select: Greedy selection with pairwise similarity cap of 0.80
- Target batch: Up to 100 candidates
The complete pre-registered rule is in docs/evidence/SELECTION_RULE.md.
The pilot panel selects the 20 best candidates from the ranked batch for first-wave synthesis.
Each candidate is scored by pilot_priority:
pilot_priority = ensemble
− 0.30 × disagreement
+ 0.05 × serum_stability
+ 0.05 × novelty
+ 0.05 × selectivity_proxy
− cytotox_penalty
cytotox_penalty = 0.0 if selectivity_proxy ≥ 0.5, else 0.05 × (0.5 − proxy) / 0.5
| Term | Rationale |
|---|---|
| −0.30 × disagreement | Penalises candidates where activity vs Boman scorers diverge |
| +0.05 × serum_stability | Rewards proteolytically stable candidates |
| +0.05 × novelty | Rewards structural distance from known references |
| +0.05 × selectivity_proxy | Rewards candidates with favourable charge/GRAVY selectivity profile |
| −cytotox_penalty | Extra demerit for HIGH_CYTOTOX_RISK tier (proxy < 0.5); max −0.05 additional |
Net selectivity swing: a perfectly selective candidate (proxy = 1.0) outranks a temporin-like low-selectivity candidate (proxy = 0.30) by 0.055 points — the same as ensemble gaining ~1.5 standard deviations in the demo pool.
Selection rules (in order):
- One representative per distinct seed (best by pilot_priority)
- Remaining slots filled by priority, subject to
max_per_seedcap (default 4) - Optional diversity filter: similarity_threshold = 0.75 between panel members
Each selected candidate receives a JSON evidence certificate (outputs/phase3_evidence/),
validated against schemas/candidate.schema.json. Certificates include:
- Candidate ID, sequence, and source
- All physicochemical features
- All individual and ensemble scores
- Nearest reference sequence and novelty distance
- Selection reason and known failure modes
- Pipeline version, config hash, and generation timestamp
All pipeline outputs are deterministic given:
- Fixed input sequences (
examples/sequences/phase3_pool.csv) - Fixed reference set (
examples/known_reference/amp_curated_references.csv) - Fixed config (
configs/phase3.yaml) - Fixed RNG seed (rng_seed=2024 in generator)
- Fixed pipeline version (
pipeline_version: 0.1.0)
The run manifest (outputs/phase3_manifest.json) records:
- SHA-256 hashes of all input files
- Config hash (SHA-256 of sorted JSON representation)
- Pipeline version
- Run timestamp
To reproduce: git checkout <commit> && make phase3
Phase 2 benchmarks verified that the pipeline:
- Recovers hidden known-positive AMPs better than random baseline (EF > 1.0)
- Remains meaningful under cluster-split evaluation (no near-duplicate leakage)
- Down-ranks poly-cationic and poly-hydrophobic negatives
- Produces stable rankings under repeated runs (reproducibility)
- Shows performance degradation when key scoring dimensions are ablated
The benchmark was expanded from 43+44 (n=87) to 95 AMPs + 96 decoys (n=191) in PR #110 to improve generalisability. 52 well-characterised public-domain AMPs were added from 12 taxonomic classes: human defensins (hBD-1/2/3, histatin 5, dermcidin, hepcidin), insect (apidaecin, drosocin, formaecin, thanatin), mammalian cathelicidins (BMAP-27/28, SMAP-29, PMAP-23/36, protegrins 1-3, PR-39), plant (RsAFP1/2, snakin-1), fungal (plectasin, eurocin), fish (pleurocidin, piscidins), marine (tachyplesins, polyphemusin, arenicin), and additional frog families. Background decoys were expanded to 96.
Retrospective AUROC — pipeline.yaml: 0.7832 (bootstrap CI₉₅: 0.72–0.84, n=191,
n_bootstrap=2000). Interpretation: STRONG (AUROC > 0.70). The lower point estimate
compared to the original demo set (0.8420) reflects that the expanded benchmark includes
diverse AMP classes (disulfide-stabilised defensins, proline-rich intracellular, lantibiotics)
that the helic-centric scorer was not calibrated for — the 0.7832 value is a more honest
and generalisable estimate of model discrimination on novel AMP classes.
Retrospective AUROC — phase3.yaml (synthesis gate): 0.7448 (bootstrap CI₉₅: 0.68–0.81,
n=191, n_bootstrap=2000). Phase3 uses re-weighted ensemble scores (activity=0.35, safety=0.30,
synthesis=0.20, novelty=0.15 vs pipeline.yaml activity=0.40, safety=0.25, synthesis=0.15,
novelty=0.20) and a stricter safety gate (max_safety_risk=0.40). Interpretation: STRONG
(AUROC > 0.70).
Important: Both point estimates comfortably exceed the AUROC > 0.70 synthesis gate. The expanded benchmark is the primary reference for model discrimination. The original demo set (43 AMPs + 44 decoys, AUROC 0.8420 pipeline / 0.8266 phase3) remains the historical baseline but should not be cited as the current benchmark in external communications.
AUPRC: 0.8164 for pipeline.yaml (+0.3190 above random baseline 0.4974);
0.7933 for phase3.yaml. Random baseline = dataset prevalence (95/191 ≈ 0.4974).
To reproduce: make validate-scoring (pipeline.yaml weights) or
make validate-scoring-phase3 (synthesis gate weights, phase3.yaml) or
make validate-scoring-strict for the composition-matched (scrambled-decoy) variant.
Key scoring changes in v0.2.x–v0.8.x (PRs #48–72):
| PR | Change | AUROC impact |
|---|---|---|
| #47 | Selectivity proxy (charge/GRAVY) | + |
| #48 | Selectivity proxy routed into pilot_priority | pilot ranking |
| #49 | Elastase resistance + aggregation propensity | + |
| #50 | Aggregation-safe mutations + balanced K/R variants | generation |
| #51 | Proline synthesis penalty (−0.10 at >15%); helix bonus 0.01→0.03 | 0.8138→0.8164 |
| #53 | Safety uses charge_density_ph74; stronger cytotox_penalty in pilot | pilot ranking |
| #58 | AUPRC added; SEED-009 (Bac2A) + SEED-010 (KR-12) added | +2 scaffolds |
| #65 | Trp-weighted aromatic bonus (1.5× vs Phe/Tyr); safety abs() fix | 0.8164→0.8086 |
| #66 | Remove duplicate REF-GIG-001 (magainin-2 counted twice in validation set) | 0.8086→0.8047 |
| #70 | Windowed mu_h (window=11, Eisenberg standard); anionic charge guard | 0.8047→0.8348 (pipeline); 0.7846→0.8126 (phase3) |
| #72 | Face segregation bonus (helix_wheel_amphipathic_score × 0.05); max_disagreement 0.40→0.45 | 0.8348→0.8420 (pipeline); 0.8126→0.8266 (phase3) — original demo set (43+44) |
| #110 | Expanded benchmark: 95 AMPs + 96 decoys (n=191); 52 new AMPs from 12 taxonomic classes | 0.8420→0.7832 (pipeline); 0.8266→0.7448 (phase3) — expanded benchmark |
PR #70 detail — windowed hydrophobic moment: For sequences > 11 residues, the full-sequence
μH is diluted by non-helical termini and linker regions. max_windowed_hydrophobic_moment()
computes the maximum μH over all 11-residue windows (the Eisenberg standard helical period).
For magainin-2 (23 AA): full-seq μH = 0.4548 vs windowed μH = 0.6862 (1.51×). For cecropin-A
(37 AA): full-seq μH = 0.2267 vs windowed μH = 0.6567 (2.90×). Short sequences (≤ 11 AA)
are unaffected: windowed == full-seq.
PR #70 detail — anionic charge guard: Anionic peptides (charge_density_ph74 < 0.0) are
electrostatically repelled by bacterial membranes (phosphatidylglycerol, cardiolipin surface)
and cannot operate via the canonical cationic AMP mechanism. The guard returns 0.0 immediately,
preventing partial credit from hydrophobic/length terms. Boundary: exactly 0.0 charge density
is not blocked (reflects neutral peptides rather than anionic). Fixed a Python eagerness bug
in feature dict lookup (features.get(key, features["fallback"]) evaluates fallback eagerly;
replaced with explicit if "key" in features conditional).
PR #72 detail — face segregation bonus: helix_wheel_faces() (PR #71) computes
face_contrast = hydrophobic_face_mean_h − hydrophilic_face_mean_h using the hydrophobic
moment vector to orient the helix wheel. amphipathic_score = max(0, min(1, contrast/2.0)).
This captures whether cationic residues are properly segregated onto the hydrophilic face — a
complementary signal to μH (which measures total moment magnitude without regard for charge
distribution). Weight 0.05 was chosen at the knee of the AUROC vs weight curve (activity-only
AUROC +0.90% vs baseline at w=0.05; +1.59% at w=0.15 — conservative choice to avoid
over-penalising non-helical AMPs). max_disagreement raised 0.40 → 0.45 in both configs because
face_segregation_bonus correctly elevated SEED-008 Trp-rich activity scores by ~0.03 score
units (their Trp clusters form a genuine hydrophobic face), pushing disagreement from ~0.40 to
~0.43. No non-Trp-rich candidate in the 709-sequence phase3 pool exceeds disagreement=0.41,
so the new 0.45 threshold still blocks genuinely uncertain candidates.
Literature: Wieprecht et al. (1997) Biochemistry 36:6124; Tossi et al. (2000) Biopolymers 55:4.
Limitation: The expanded benchmark (95 AMPs + 96 decoys, n=191) includes diverse AMP classes from 12 taxonomic families but is still far from the scale of curated databases (APD3 > 3,000 AMPs, DRAMP > 19,000 entries). Benchmark expansion to >500 sequences with composition-matched decoys and cluster-split evaluation is identified as deferred work (v1.0+) in ROADMAP.md. Current AUROC estimates should be interpreted as a lower bound on discrimination: the model is tested against diverse AMP classes it was not calibrated for, which is a stricter test than a compositionally homogeneous benchmark.
| Failure mode | Impact | Status |
|---|---|---|
| Scoring not validated against lab data | Activity/safety scores may not correlate with assay results | Known; lab calibration needed |
| Near-seed generation only | Misses genuinely novel AMP families | Known; future work |
| Small reference set | Novelty may be overestimated | Partially mitigated; 72 refs used |
| Two scorers only | Disagreement is coarse uncertainty; not a calibrated posterior | Partially mitigated; Boman index added in v0.2 |
| Helical amphipathicity assumption | Non-helical AMPs underscored | Known limitation |
| No structural modeling | Cannot predict membrane interaction mode | Out of scope v0.1 |
- Bieth, J.G. (1986). Elastases: structure, function and pathological role. Bull Eur Physiopathol Respir, 22(Suppl 7), 7–12.
- Bjellqvist, B. et al. (1993). The focusing positions of polypeptides in immobilized pH gradients can be predicted from their amino acid sequences. Electrophoresis, 14(10), 1023–1031.
- Boman, H.G. (2003). Peptide antibiotics and their role in innate immunity. Annual Review of Immunology, 13, 61–92.
- Chou, P.Y. & Fasman, G.D. (1974). Conformational parameters for amino acids in helical, beta-sheet, and random coil regions calculated from proteins. Biochemistry, 13(2), 222–245.
- Dathe, M. & Wieprecht, T. (1999). Structural features of helical antimicrobial peptides: their potential to modulate activity on model membranes and biological cells. Biochim Biophys Acta, 1462(1–2), 71–87.
- Doherty, J.B. et al. (1991). Potent and selective human neutrophil elastase inhibitors. Biochemistry, 30(35), 8540–8550.
- Eisenberg, D. et al. (1984). Analysis of membrane and surface protein sequences with the hydrophobic moment plot. J Mol Biol, 179(1), 125–142.
- Fields, G.B. & Noble, R.L. (1990). Solid phase peptide synthesis utilizing 9-fluorenylmethoxycarbonyl amino acids. Int J Pept Protein Res, 35(3), 161–214.
- Hancock, R.E. & Sahl, H.G. (2006). Antimicrobial and host-defense peptides as new anti-infective therapeutic strategies. Nature Biotechnology, 24(12), 1551–1557.
- Hilpert, K. et al. (2006). Short cationic antimicrobial peptides interact with ATP. Antimicrob Agents Chemother, 50(12), 4181–4188.
- Huang, H.W. (2000). Action of antimicrobial peptides: two-state model. Biochemistry, 39(29), 8347–8352.
- Jenssen, H. et al. (2006). Peptide antimicrobial agents. Clinical Microbiology Reviews, 19(3), 491–511.
- Kyte, J. & Doolittle, R.F. (1982). A simple method for displaying the hydropathic character of a protein. J Mol Biol, 157(1), 105–132.
- Tossi, A. et al. (2000). Amphipathic, alpha-helical antimicrobial peptides. Biopolymers, 55(1), 4–30.
- Wade, D. et al. (1990). All-D amino acid-containing channel-forming antibiotic peptides. Proc Natl Acad Sci USA, 87(12), 4761–4765.
- Wimley, W.C. & White, S.H. (1996). Experimentally determined hydrophobicity scale for proteins at membrane interfaces. Nat Struct Biol, 3(10), 842–848.
- Zasloff, M. (1987). Magainins, a class of antimicrobial peptides from Xenopus skin. PNAS, 84(15), 5449–5453.
- Zasloff, M. (2002). Antimicrobial peptides of multicellular organisms. Nature, 415(6870), 389–395.
Full candidate-level references are in each evidence certificate.
This pipeline does not optimize for toxicity, virulence, immune evasion, or any harmful objective. All candidates are short, linear, non-targeted peptides evaluated for potential antimicrobial utility only. No dangerous organism protocols, pathogen enhancement, or harmful objectives are present in this repository. Generator outputs have not been screened against bioweapon databases; expert review is required before any external release.