Dockquest

"Claude Code for protein–ligand docking"

Agentic • Automated • Monetizable

doi.bio Agentic DevTools Canada

Steven Ness • sness@sness.net • dockquest.doi.bio

The Problem

  • Drug discovery is slow, expensive, failure‑prone (10–15 years; $2.6B; 90% fail).
  • Computational tools are being developed rapidly
  • Many existing tools and algorithms exist, hard to know what to use and how to use it.
  • Much information on how to use these tools and interpret results are locked up in papers.

Team

  • Steven R. Ness Ph.D. (papers)
  • Background in Software Engineering, Machine Learning, Structural Biology, Rational Drug Design
  • CWSF Designer Genes : DockVision : CRANK
  • Participated in CASP2, CASP15, CASP16
  • Worked as molecular modeller designing drugs

Rational Drug Design

  • Pick a validated target (protein) tied to the disease →
  • Get its 3D shape (experiment or AI models) →
  • Design/score molecules for best “fit” and properties →
  • Iterate quickly with wet-lab feedback.

Vision

  • Bring new AI tools to drug design
  • Help scientists use these new AI tools
  • Disrupt all existing drug design software with an agentic model

Moat

  • DockQuest Glue + Provenance Store — Hard to rebuild and becomes workflow bedrock.
  • doi.bio Literature Graph — Curated paper→residue/ligand links; unique, compounding data asset.
  • Learned Reranker + Explainability — Model that re-scores poses using gallery + literature features, producing trusted, regulator-friendly narratives.

Our Insight

  • Claude Code pattern is powerful
  • Structural Biologists live in CLI and filesystem
  • Claude Code lives in the CLI and filesystem
  • Run agents where the scientists live
  • Give the agents access to the tools, datasets and scientific papers

Solution — What DockQuest Does

  • Fold: ESM3/AF‑class structure hypotheses
  • Prep: RDKit ligands • protein prep
  • Dock: GNINA / AutoDock‑GPU with CNN scoring
  • Relax: OpenMM quick minimization
  • Explain and Refine: doi.bio literature retrieval
  • Reproduce: Inputs, parameters and results saved

CLI‑first; optional web viz (3Dmol.js); MCP/IDE integration; air‑gapped packaging.

Current tools

  • Many tools are CLI tools and use the unix filesystem
  • Dockquest lives in the unix filesystem and CLI
  • Communicate with cloud servers to run jobs in the cloud

Demo

% dockquest
terminal

Live 3D — Pose & Pocket

viewer

Interactive 3D (3Dmol.js) with surface/pocket highlights, pose gallery, and citations.

Why Now

  • ESM3/AF‑class unlocked folding; missing layer is operational glue.
  • Vast amount of untapped datasets and scientific papers unlocked
  • Apply ideas from Claude Code/Codex/Gemini to Docking

Market

  • TAM: $8-12B (SMB biotechs, CROs, academic cores, early pharma).
  • SAM: $2.5-5.5B (protein-ligand focus small molecules).
  • SOM (5‑yr): ~$50–500M via agentic orchestration + OEM/API.

Business Model & Pricing

  • Academic (free + usage)
  • Industry Researcher ($99/mo + usage)
  • Enterprise (Custom) (~$50K–2M/yr)
  • Usage‑metered compute credits • OEM/API licensing • MCP Server

Roadmap (12 Months)

  • 0–3 mo: v0.9 agent/adapters/reports • 3 LOIs
  • 4–6 mo: MCP Server • Pilots 1–3 live • first conversion
  • 7–12 mo: 5 pilots → 2 enterprise • 2 OEM trials • CASP17

Ask (Pre‑Seed)

$500k for 12 months

  • 70% GTM‑Science ($350k): 2 Scientist‑BDs
  • 20% Compute & Data ($100k): GPUs/CPU, storage, pilots
  • 10% Trust & Ops ($50k): legal, SOC2 prep

Contact

Steven Ness • sness@sness.net • dockquest.doi.bio