Overview
Titer takes a single-letter amino acid sequence and returns a typed dossier:
- →Identity · canonical sequence, SHA-256 fingerprint, grade A→X, source citation
- →Structure · AlphaFold v6, ESMFold or PEP-FOLD4 model routed by length
- →Mechanism · ChEMBL + GtoPdb + OpenTargets, synthesised by gpt-5-2 with cited claims re-verified against their sources
- →Batch + COA · vendor bond status, lab-signed COA hash, mass-spec hash binding
- →Receipt · two Solana memos (commit + reveal), Arweave CID for the full dossier
The free tier uses a Titer paymaster to sponsor on-chain writes. Connect a wallet to self-pay, save lookup history, or unlock the cohort outcome layer.
Pipeline
Five stages, p50 1.4s for the identity card and p95 4.8s for the full dossier. Identical canonical sequences hit the cache and short-circuit any stage that has already resolved.
01 · canonicalize parse → strip whitespace/numbering →
normalize (U→C, O→K) → SHA-256
02 · identity seed table → UniProt PS exact (≥7aa) →
BLAST homolog (e ≤ 1e-10) → grade
03 · structure if length < 40: PEP-FOLD4
if 40-400 + UniProt: AlphaFold DB v6
if 40-400 + no hit: ESMFold
if length > 400: AlphaFold DB only
04 · batch + COA vendor bond status · lab-signed COA hash
mass-spec hash binding (titer.ms.v1)
05 · receipt commit memo (~400ms confirmed)
dossier upload to Arweave
reveal memo (salt + result CID + result hash)Grading rubric
Each dossier carries one of five letter grades. The grade encodes confidence in the sequence identity only. It does not encode safety, efficacy, or vendor quality (those are separate primitives).
| Grade | Threshold | Meaning |
|---|---|---|
| A | Seed table or UniProt PS ≥ 99% | Exact match |
| B | BLAST ≥ 90% identity | Close homolog |
| C | BLAST ≥ 70% identity | Distant homolog |
| D | no reliable hit | Novel sequence |
| X | failed canonicalization | Invalid input |
Commit / Reveal
Each lookup produces two transactions on Solana mainnet-beta. The commit lands at confirmed in ~400ms, before any third-party API runs. The reveal lands after the dossier is uploaded to Arweave.
// commit memo titer:commit:v1:<sha256(input || salt || requested_at)>:<requested_at> // reveal memo titer:reveal:v1:<commit_tx_sig>:<salt>:<result_cid>:<result_hash>
The commit hash is computed over a canonical JSON message that includes the input, the optional target class, the schema string "titer.commit.v1", the salt, and the requested-at timestamp. Anyone with the salt can re-hash and verify the run was honest end-to-end without trusting Titer’s servers.
Mass-spec hash
The lab uploads the raw spectrum (mzML / mzXML). Titer computes a content-stable hash by binning peaks at 0.01 Da, normalizing to the max peak, dropping anything below 0.1% relative intensity, and SHA-256 over the sorted bin:intensity entries.
schema titer.ms.v1 binning bin = round(mz / 0.01) norm intensity / max_intensity threshold 0.001 (drop below 0.1% relative) canonical sorted "bin:intensity" entries, comma-joined hash sha256(canonical) memo titer:ms:v1:<batch_id>:<sha256>:<lab_id>:<acquired_at> fallback cosine similarity over binned vectors when hashes differ
Hardware-to-hardware drift below the bin width is absorbed. Different compounds on the same instrument do not collide. When two labs scan the same compound on different instruments, the cosine fallback (over the same binned vectors) carries the signal.
Vendor bond
Vendors stake $TITER to list. Bond size scales with cumulative batch GMV and is multiplied by the vendor’s historical incident rate (capped at 4×).
required_bond_lamports(gmv, incident_rate):
base = ceil(gmv * 0.05)
multiplier = 1 + min(incident_rate * 4, 3)
return max(5_000_000_000, ceil(base * multiplier)) // floor 5 SOL
insurance_premium_lamports(vial_price, quality_score):
base = ceil(vial_price * 0.02) // 2% default
discount = clamp(quality_score, 0, 0.5)
return ceil(base * (1 - discount))Disputes
A buyer may file a dispute against a batch with: (i) an independent mass-spec hash from any partner lab, (ii) the cosine similarity to the vendor-claimed reference spectrum, (iii) photo evidence, (iv) a free-form description.
Filed disputes lock the bond pending arbitration. Three token-staked arbitrators are randomly selected; majority decision settles the case. Slashes pay out to impacted buyers from the same pool. The full evidence trail is committed on chain.
Cohort layer
Buyers log N-of-1 outcomes against a verified batch ID. Logs are pseudonymous and processed in the browser by default · biomarker uploads stay in encrypted IndexedDB unless the user opts in to aggregate publication.
| N (logs per peptide) | Cohort grade | Reading |
|---|---|---|
| ≥ 1000 | A | Controlled signal |
| ≥ 200 | B | Preliminary |
| ≥ 50 | C | Directional |
| < 50 | D | Insufficient evidence |
Tracked metrics include subjective efficacy, side-effect severity, recovery time, biomarkers (hsCRP, HbA1c, IGF-1, free T, SHBG), and body composition. The full schema is in src/lib/cohort.ts.
Data sources
| Source | Use | License |
|---|---|---|
| UniProt | Identity match | CC-BY 4.0 |
| AlphaFold DB v6 | Structure (40–400aa with UniProt hit) | CC-BY 4.0 |
| ESMFold | Structure (40–400aa, no UniProt hit) | MIT |
| PEP-FOLD4 | Structure (< 40aa) | Academic free |
| ChEMBL release 35 | Mechanism, ligand binding | CC-BY 4.0 |
| GtoPdb (IUPHAR) | Mechanism, ligand–receptor map | CC-BY 4.0 |
| OpenTargets | Mechanism, disease association | CC0 |
| Reactome | Pathway annotation | CC-BY 4.0 |
| PubMed (E-utilities) | Citations | Public |
| dbSNP / gnomAD | Variant frequency | Public |
API
| Method | Path | Purpose |
|---|---|---|
| POST | /api/lookup | Run a sequence lookup. Returns the typed dossier + commit memo. |
| POST | /api/verify | Submit a peak list, get the spectrum hash + binding memo. |
| POST | /api/commit | Anchor a commit memo to Solana (paymaster-sponsored on free tier). |
| GET | /api/batch/:id | Fetch a batch by ID · bond, COA, MS hash, dispute status. |
| GET | /api/vendor/:id | Fetch vendor profile · bond, incident rate, quality score. |
| POST | /api/cohort/log | Submit an N-of-1 outcome log (wallet-signed). |
| GET | /api/cohort/:peptide | Aggregate cohort rollup for a peptide. |
Roadmap
| Phase | Status | Scope |
|---|---|---|
| 1 · Lookup core | shipped | Canonicalize · UniProt · AlphaFold · grading · commit/reveal |
| 2 · MS hash + batches | in progress | Lab signing · vendor bonds · batch ledger · embed badge |
| 3 · Cohort layer | next | N-of-1 logging · cohort grades · biomarker schema |
| 4 · Disputes | next | Token-staked arbitration · recall coordination |
| 5 · Reasoning | next | Cited-claim verifier · reproducibility budget · adversarial re-runs |
| 6 · Privacy | later | ZK researcher credentials · Arcium MXE for private BLAST |
Legal posture
Research only. Not for human use. No medical claim is made or implied anywhere on this site or in any output.
Titer does not sell, broker, or recommend vendors. The vendor table shows price, COA freshness, bond, and dispute history · never a “best of” or “buy now” rail. Independent COAs are linked for reference; verify with your own lab before any use.
Vendor scrapers identify honestly (User-Agent Titer-VendorBot/1.0), respect robots.txt, throttle, and never republish vendor content verbatim · only price + outbound COA link. 18+. DMCA agent registered. 21 CFR aware.