RDKit compound triage: drug-likeness, PAINS and Brenk
Validate and standardize SMILES, compute Lipinski/Veber/QED, flag PAINS and Brenk alerts, and get a pass/review/fail verdict per compound.
Free
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsRDKit compound triage: drug-likeness, PAINS and Brenk
Example session with this skill installed
Triage the compounds in hits.csv and tell me which are worth following up.
- Read your context and instructions
- Compiled the rdkit compound triage
- Generated the data export
Script summary: Triaged 5 compounds: 2 pass, 1 review, 1 fail, 1 invalid SMILES.
| id | verdict | MW | cLogP | QED | alerts |
|---|---|---|---|---|---|
| benzylidene_rhodanine | fail | 221.3 | 2.18 | 0.58 | PAINS:ene_rhod_A; Brenk:Michael_acceptor_1; Brenk:Thiocarbonyl_group |
| aspirin | review | 180.2 | 1.31 | 0.55 | Brenk:phenol_ester |
| caffeine | pass | 194.2 | -1.03 | 0.54 | none |
| imatinib | pass | 493.6 | 4.59 | 0.39 | none |
| bad_entry | invalid | SMILES did not parse |
Alerts explained
ene_rhod_A (PAINS): 5-benzylidene rhodanine, a Michael acceptor and frequent assay-interference chemotype. This alone fails it.
phenol_ester (Brenk): a labile aryl ester. For aspirin that hydrolysis is the mechanism, so it is flagged for review, not rejected.
Worth following up: caffeine and imatinib (no alerts, no Ro5 violations). Fix or re-check the bad_entry SMILES.
Full table: triage.csv
triage.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Give your agent a SMILES list or a CSV of hits and get a med-chem triage report back.
What it does
- Standardize: RDKit cleanup, salt stripping and neutralization. Invalid SMILES are reported, not silently dropped.
- Properties: MW, cLogP, HBD/HBA, TPSA, rotatable bonds, QED.
- Rules: Lipinski Ro5 violations and Veber criteria.
- Alerts: PAINS (assay interference) and Brenk (unwanted or reactive groups).
- Verdict: pass / review / fail per compound, plus a full CSV.
Why it's trustworthy
Every number comes from a bundled, self-tested RDKit script (scripts/triage.py --selftest); the model never improvises a descriptor. The agent then explains each alert in medicinal-chemistry terms and reminds you that alerts are prompts for a chemist, not automatic rejections.
Requirements
uv. RDKit installs automatically on first run. Accepts CSV, .smi/.txt or a single SMILES string.
- Rule-based filters (Lipinski, Veber, PAINS, Brenk) for prioritization; they flag compounds for a chemist to review and are not a verdict on activity or safety.
- Many approved drugs (natural products, macrocycles, beyond-Rule-of-5) violate Lipinski.
- Needs uv and one-time internet access to install RDKit.
- Single-structure standardization only: no tautomer enumeration or 3D/conformer work.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 28 days ago
- Free with an Agensi account
Requires
Astral
PythonPermissions
Allowed Hosts
Shell: runs the bundled scripts/triage.py with uv. Network: only uv downloading RDKit on first run. Read/Write: reads your input CSV/.smi and writes the triage CSV you name with -o.