01100%
Reference match
LL-37 resolves to a cathelicidin library hit with exact sequence coverage.
DeSCI - making science accessible
AI peptide lab
terminal.
We’re entering a new era of peptide discovery — where frontier AI can turn deeper molecular insight into faster candidate decisions, bringing autonomous lab intelligence closer to new treatments that matter.
Everyone can be a scientist
The goal: build the largest crowdsourced peptide candidate database.
Driven by autonomous agents, every submitted sequence becomes an analyzed, scored, and receipt-backed candidate in a growing peptide intelligence network. Brought on-chain through verifiable research receipts — open database can be used by researchers worldwide for novel drug discovery and treatment candidates. Each Solana Memo acts as a timestamped contribution claim — an open trail researchers can credit, inspect, and build on.
01Contribution loop
Submit peptide ideas, generate receipt-backed candidate records. If a researcher later explores that candidate, cites it, publishes around it, or builds IP from it, they can credit the original contributor.
02Quality / token loop
More trusted contributors, better data. $PepOS expands launch-phase run capacity while base submissions and browsing stay open. Reward and credit mechanics come later as receipts build the contribution graph.
03Intelligence loop
Every analyzed candidate expands PepAgent’s signal map with more examples, edge cases, motif patterns, and scoring outcomes.
Why it matters
From peptide idea to drug-discovery candidate record.
PepAgent turns a raw sequence or peptide name into a structured triage record: what it resembles, which biochemical signals look strong, which weak signals should be dropped, and what should be handed off for deeper research.
The point is credible early-stage narrowing: score the signal, expose the caveats, preserve the record, and move only the best candidates toward deeper modeling, synthesis planning, or wet-lab review.
PepAgent triage mapFrom molecular signal to candidate decision
receipt-backed handoff02+6
Biochemical profile
Charge, hydrophobicity, motif, and amphipathicity signals are split into explainable evidence lanes.
031 flag
Triage flags
Risk and follow-up signals are surfaced before a candidate enters the research queue.
04SHA-256
Receipt record
The run becomes a receipt-backed candidate card with a wallet-signable Solana Memo payload.
Research queueDecision layer
LL-37Promoteknown AMP · high cationic signal
MelittinWatchmembrane-active · hydrophobicity caution
WFWFWKKExplorenovel sequence · no library match
poly-RDroprepeat-dominated · low complexity
OutputCandidate card + flags + SHA-256 receipt
scoredropwatchhandoff
How PeptideOS works
PepAgent designs, analyzes, scores, and logs candidate peptide ideas through a seven-stage AI workflow — then anchors the result as a SHA-256 receipt on Solana.
⬡
Specialised AI PepAgent
PepAgent designs, analyzes, and scores peptide ideas through transparent biochemical gates, reference matching, motif signals, and receipt-backed run records.
◈
Peptide reference library
Cross-referenced sequences spanning antimicrobials, cathelicidins, defensins, cell-penetrating peptides, neuropeptides, hormones, and cyclic research candidates. Family assignments, motif patterns, and characterisation status logged per entry.
◉
Solana-anchored receipts
Each run produces a SHA-256 receipt hash and a UTF-8 Memo payload. Connected wallets can anchor receipts to Solana mainnet via the Memo Program — no custody, explicit approval.
◌
Research boundaries enforced
Gates, flags, and copy are scoped to research informatics. No diagnosis, therapeutic, or clinical validation claims. Flags surface when candidates fall outside characterised territory.
Analysis engine
Seven-stage specialised AI peptide workflow — scored, logged, and receipt-anchored.
01
Design
Intake peptide name, sequence, or candidate direction — prepare a PepAgent run context.
Accepts common names, reference identifiers, or raw IUPAC sequences
02
Parse
Classify input, resolve aliases and reference names, separate sequence vs candidate intent.
Resolves cathelicidin, defensin, neuropeptide names to canonical sequence form
03
Normalize
Clean residues, validate amino-acid alphabet, compute length and composition baseline.
20 standard amino-acid gate; short · medium · long band classification
04
Match
Compare against known peptide reference signals, motifs, families, and sequence similarity.
Antimicrobials · cathelicidins · defensins · CPPs · neuropeptides · cyclic peptides
05
Score
Run biochemical gates: charge, hydrophobicity, amphipathicity, cysteine bridge potential, novelty.
Specialised AI PepAgent triage with transparent biochemical gates
06
Flag
Surface cationic, hydrophobic, cysteine, proline, and novelty signals with research-boundary cautions.
10 flag types; candidates outside characterized territory are explicitly marked
07
Receipt
Compile canonical run fields into SHA-256 receipt hash and Memo-ready lab payload.
Identical inputs always produce identical hashes; anchor via Memo Program, explicit approval
Library candidates
View all →antimicrobial
Melittin
GIGAVLKVLTTGLPALIS…
Bee-venom amphipathic AMP; classic membrane-disruption model.
View analysis →
antimicrobial
Magainin-2
GIGKFLHSAKKFGKAFVG…
Frog-skin toroidal-pore AMP; amphipathic alpha-helix scaffold.
View analysis →
cathelicidin
LL-37
LLGDFFRKSKEKIGKEFK…
Only known human cathelicidin AMP; cationic and immunomodulatory.
View analysis →
cell-penetrating
Penetratin
RQIKIWFQNRRMKWKK
First reported CPP; from Antennapedia homeodomain helix 3.
View analysis →
In the lab
Example runs showing how PepAgent parses, matches, and flags different input types.
Query
LL-37cathelicidinmediumHIGH_CATIONIC
Name resolved → sequence. Cationic flag expected for human cathelicidin AMP.
Query
GIGAVLKVLTTGLPALISWIKRKRQQantimicrobialshortHIGH_CATIONICHIGH_HYDROPHOBICITY
Raw sequence matched Melittin exactly. Amphipathic α-helix pattern confirmed.
Query
ACYCRIPACIAGERRYGTCIYQGRLWAFCCdefensinshortODD_CYS
HNP-1 defensin. Odd cysteine flag from triple disulfide accounting edge.