skills.rest
openai
openaiOFFICIAL
Explore all skills by this author →
WebsiteGitHub

boltz-small-molecule-design

Design novel small-molecule binders for protein targets using the Boltz API.

5.3k|765|Updated Mar 4, 2026
Search Tags:#drug discovery#boltz#adme#binding confidence#protein target#ligand generation#small molecule design
One-click install
npx skills add https://github.com/openai/plugins --skill boltz-small-molecule-design
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: boltz-small-molecule-design
Source: https://github.com/openai/plugins/tree/main/plugins/boltz-api-cli/skills/boltz-small-molecule-design
Command: npx skills add https://github.com/openai/plugins --skill boltz-small-molecule-design

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Generating new drug-like molecules that bind a specific protein target normally requires specialized computational chemistry infrastructure. This Skill drives the Boltz small-molecule design API end to end, from payload authoring and cost estimation through job submission, background result download, and hit ranking.

Core Features & Use Cases

  • De Novo Binder Generation: Submit protein sequences with optional pocket residues and reference ligands to generate 10 to 1,000,000 novel candidate molecules.
  • Cost Control and Confirmation: Run estimate-cost before submission, quote the flat $0.025 per molecule rate, and wait for explicit user confirmation before spending.
  • Filtering and Ranking: Apply optional molecule filters (Lipinski, RDKit descriptors, PAINS/SMARTS catalogs) and rank results by binding_confidence or optimization_score with free ADME triage data.
  • Use Case: A computational chemist provides a target protein sequence and known binder SMILES, confirms a $250 estimate for 10,000 molecules, and receives a ranked index.jsonl of generated hits with predicted structures and ADME properties.

Quick Start

Use the boltz-small-molecule-design skill to estimate the cost and design 100 novel binders for this protein sequence, then rank the results by binding confidence.

Frequently Asked Questions about boltz-small-molecule-design

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I design novel small-molecule binders for a protein target?▼

Build a payload with your protein sequence in target.entities, optionally add 0-based pocket_residues and reference_ligands, set num_molecules, then run boltz-api small-molecule:design estimate-cost and start. Download results with boltz-api download-results and rank hits from results/index.jsonl.

How much does Boltz small molecule design cost?▼

Cost is a flat $0.025 per generated molecule, independent of target or molecule size. Always run estimate-cost first and report the estimated_cost_usd value as the authoritative total before submitting the job.

What is the minimum number of molecules for Boltz design jobs?▼

The server rejects num_molecules below 10 or above 1,000,000 with a validation error. Validate the range client-side before calling estimate-cost, and propose 10 if the user requests fewer.

Can I restrict generated molecules to synthesizable compounds?▼

Yes, set chemical_space to enamine_real to restrict generation to synthesizable molecules from the Enamine REAL library. Only add this option when the user explicitly wants generation constrained to that library.

How do I rank Boltz design results for hit discovery vs lead optimization?▼

Sort results/index.jsonl by binding_confidence for hit discovery or by optimization_score for lead optimization. Each molecule also includes a free ADME block with solubility, permeability, and lipophilicity for developability triage.

When should I not use de novo small molecule design?▼

Do not use it for screening an existing compound library or one-off docking of known molecules; those fit library screening endpoints instead. It is intended for generating novel ligands against a target without a fixed compound collection.

Related Skills

Explore More Education & Research Skills→
learningmatter-mitlearningmatter-mit
144

drug-binding-site-definition

Compute docking box center and dimensions from ligand, residues, or JSON.

Official
Intermediate
jakechen1jakechen1

adme-property-predictor

Predict ADME properties and drug-likeness scores from SMILES strings.

Community
Intermediate
001TMF001TMF
104

pxdesign

Generate de novo protein binders against target structures via CLI workflows.

Community
Advanced
InternScienceInternScience
28

molclaw-equiscore-docking

Rank candidate molecules by docking score against a protein target.

Official
Advanced