doi.bio

The connected,
AI-native graph
of all biology.

Proteins link to structures. Papers link to researchers. Everything connects — already wired, ready for the machines that now read science.

Pre-seed · Raising $750K Steven Ness, PhD  ·  sness@sness.net
The Problem

Biology's knowledge is vast — and disconnected.

Drug discovery runs on 35M+ papers, 250M+ proteins and 220K+ structures that live in separate silos with no shared wiring.

"Data sources vary in formats, metadata and identifiers, lacking standardization — making it difficult to merge information across fields."— Taylor & Francis, 2025
  • A scientist asks "who worked on this target, what did they try, what bound?" — and burns days clicking between PubMed, PDB, UniProt, Scholar & LinkedIn.
  • Every AI drug-discovery team rebuilds the same knowledge graph from scratch — costing FTEs, never staying current.
  • Incumbents (Clarivate, CAS) were built for humans browsing a webpage — not the LLMs and agents now doing the reading.
The data is free. The connections are not.
Why Now

A three-year window just opened.

  • AI agents are the new readers of science. They need pre-linked, pre-structured corpora — not HTML built for human eyes.The consumer of scientific knowledge changed. The format hasn't.
  • The structural-biology data explosion — AlphaFold, PDB, UniProt, PMC full-text — finally makes a complete graph possible.
  • Every techbio company is building this internally — and hating it. The build-vs-buy line just moved.

The window to become the default substrate — before each lab's in-house graph calcifies into permanence — is open now.

Be the layer
before the lock-in.
The Solution

A graph you traverse — not a database you query.

A pre-connected knowledge graph of biology, stored as plain markdown + YAML — the native diet of LLMs. Protein → structure → paper → author → lab → company, already wired.

35M+
Papers (PubMed)
8M+
Full-text articles (PMC)
250M+
Proteins (UniProt)
226K+
Structures (PDB)
67K+
Researcher profiles
104K+
Unified person records
1,000+
Biotech companies
114K+
Files, all plain-text
The Hard Part — Already Done

The connections already exist.

Entity resolution — "is this the same researcher under three name spellings?" — across 104K people. Sub-second search over 7.27M full-text articles.

Competitors have the data.
They don't have the wiring.
PDB structure → UniProt protein290,641
PDB structure → PMID paper204,213
Scholar → publications57,188
Protein → AlphaFold model53,171
Scholar → ORCID6,525
Indexed entities, one SQLite graph133,626
The Wedge Product

Researcher Intelligence: "Who knows about this?"

Every customer segment asks the same question. One narrow, valuable, shippable API on top of the graph — v1 in 2–4 weeks, because the graph already exists.

GET /target/{gene}/experts
→ researchers · labs · current company
   · collaborators · contact pathways
  • AI drug-discovery teams — ground model predictions in what's already been tried.
  • TechBio BD & sales — find the 15 scientists at Pfizer working on CD20, not a generic purchased list.
  • Investors & scouts — map a field's real experts in minutes, not weeks.
  • Land, then expand — the API is the front door to the full stack.
The Full Stack

A closed loop: KNOW → CREATE → TEST.

doi.bio is the knowledge layer of a three-product loop. Knowledge informs design; design feeds validation — and every result flows back into the graph.

No point-solution competitor
has the closed loop.
KNOW
doi.bio
Pre-connected knowledge graph
▼
CREATE
smol
AI molecule design
▼
TEST
DockQuest
AI-native docking & validation
▼  results feed back
The Moats

Why this compounds — and can't be cloned.

01 Core
The compounding link graph
Anyone can download PubMed. Nobody cheaply reproduces 134K resolved entities + ~800K curated links. Each new source multiplies existing connections — value grows faster than size.
02 Timing
AI-native substrate
Stored as markdown/YAML — what agents read directly. For incumbents, "be AI-native" is a re-platform that fights their own customers.
03 Founder
Founder-market fit
CASP-validated scientific judgment + the ability to build the whole data stack solo. Most knowledge-graph founders have one, not both.
04 Model
Counter-positioning on price
$100/mo, locally-hostable, vs $50K/seat. Incumbents can't match without destroying their own margin — the disruptor's wedge.
05 Lock-in
The KNOW→CREATE→TEST loop
Three integrated layers. Switching cost compounds; every docking run makes the customer's own graph better.
06 Network
Open & distributed
Every vault is a forkable git repo — kills the enterprise lock-in objection and seeds a contributor flywheel.
Founder-Market Fit

The receipts.

Steven Ness — PhD, Computer Science. Building doi.bio solo: ingestion, entity resolution, indexing and a 7.27M-article search engine.

Builds the data infra
— and knows what's real.
CASP15

Ranked #1 against all 26 competing groups on target T1124 — a blind, refereed, world-level structure-prediction benchmark.

CASP16

Found and fixed a silent input-handling bug in the canonical chai-lab reference pipeline that the rest of the field ran straight past.

SOLO

Built the entire ingestion + linking + indexing stack — 114K+ files, sub-second search — alone.

Market & Business Model

A $1B serviceable market — 70%+ margin.

TAM drug-discovery SW + sci. databases$25B+
SAM AI drug discovery + knowledge platforms~$1B
SOM Year 1–2 · 10–20 customers$0.5–1M

Land with the Researcher Intelligence API → expand into the full stack. The corpus is already built — margin is structural.

API access$99–999 / mo
Enterprise & data licensing$25–150K / yr
Embedded partner integrationsrev-share
The Ask

Raising $750K pre-seed.

18 months of runway to turn a working knowledge graph into a fundable seed-stage company.

Target: $200K+ ARR · 50 customers · 5K free users
1
Ship the wedgeLaunch the Researcher Intelligence API; sign 5 paying design partners.
2
Scale & freshnessHarden ingestion → 1M+ researchers with continuous, automatic freshness.
3
Close the loopWire doi.bio into DockQuest so KNOW→TEST is live and demoable.
doi.bio

The data is free.
The connections
are the company.

And they compound.
$750K Pre-seed Steven Ness, PhD  ·  sness@sness.net
doi.bio  /  pre-seed
← → or space to navigate
01 / 12