Cut days of drug-discovery data wrangling to about an hour: give a chemistry AI agent Cheminformatics Skills so it curates compounds from literature, runs matched molecular pair analysis, trains potency models, and proposes synthesis routes you still validate in the lab: AI agent prompt
Google DeepMind's Nov 3, 2026 blog introduces Cheminformatics Skills for Google Antigravity. In a walkthrough on lung-cancer tyrosine kinase inhibitors (crizotinib to lorlatinib-style rigidification), an agent searches literature for thousands of compounds, filters noise such as buffers, builds an interactive viewer, runs matched molecular pair analysis, trains potency models with scaffold or chronological splits, and proposes new macrocycles with synthetic routes and predicted potency--work the authors say used to take multiple days for one iteration, done in about an hour in the demo. Skills ground calculations in RDKit with ChEMBL and PubChem connectors. The post stresses that designs are computational hypotheses that must be synthesized and validated experimentally, not finished drugs.
The prompt
Act as a cheminformatics research partner for a medicinal-chemistry loop. Goal: [e.g. rigidify a floppy inhibitor for target X]. Starting molecules or papers: [names or links]. Tools I can use: [RDKit, ChEMBL, PubChem, local files]. 1) Search and curate related compounds; drop obvious non-actives like solvents or buffers; cite every source. 2) Summarize the set: count, potency ranges if present, and any clear trend (for example fewer rotatable bonds vs higher potency). 3) Run or outline a matched molecular pair analysis for the change I care about ([e.g. cyclization / macrocyclization]) and say what the pairs actually show--no invented numbers. 4) Propose how to train a simple potency model (features, split type, how I'll check it) and what would make the model untrustworthy. 5) Suggest up to 5 new designs with a one-line chemistry rationale and a plausible synthesis sketch from documented reaction types. Label every design as a hypothesis for lab validation. Never claim a compound is a drug; never invent assay data I didn't provide.
Fill in the blanks, then copy
This prompt has 4 [bracket]s to fill in. Type your details below and the prompt updates as you go. Your answers stay in this browser and are reused on other prompts with the same blanks. Nothing is sent to us.
Example values are fictional, just to show how a filled-in prompt reads. Replace them with your real details.How to use it
- Copy the prompt above.
- Replace every [bracket] with your own details (use the fill-in-the-blanks box above, which remembers your answers for the next prompt).
- Paste it into Google Antigravity with Cheminformatics Skills, RDKit, ChEMBL and review the answer before you use it.
Works with: Google Antigravity with Cheminformatics Skills, RDKit, ChEMBL, PubChem, Claude or Gemini
Real-world example: deepmind.google/blog/agentic-cheminformatics-with-google-antigravity/
Before you use it: what to check
Already built into this prompt
- Never claim a compound is a drug; never invent assay data I didn't provide.
Check yourself
- Read and edit the draft before it reaches a customer or goes public. If you later automate this prompt, start with "draft and notify me", not "send automatically".
- Check names, dates, numbers and any claim against your own source material before you use the answer. AI can misread or invent details.
Using an AI agent? Download this prompt as a SKILL.md agent skill with these checks built in, or get the .zip for Claude.ai. How to install skills.
Want to make this prompt even better? Grade it with the free prompt grader or build your own with the AI prompt generator.
32 copy-paste agent prompts (PDF) plus 5 new prompts every week.
Get it free