Chemical Literature and Patent Mapping

Goal

To programmatically check if a specific chemical compound exists in recent literature or patent databases. This skill uses PubChem's PUG-REST XRefs endpoint to extract an exhaustive list of associated PubMed IDs and Patent numbers.

This is incredibly useful as an autonomous "novelty check" for generated molecules.

Instructions

1. Extract Literature and Patents by CID

Provide the CID of the target molecule. By default, the script will output the absolute total number of hits but limits the JSON save array to 1000 to prevent memory flooding for ubiquitous molecules (like Aspirin, which has over 100,000 patents). Adjust --limit as needed.

# Env: base-agent
python .agents/skills/general-chemical-literature/scripts/get_xrefs.py \
  --cid 2244 \
  --limit 50 \
  --outdir research/aspirin_literature \
  --output xrefs_aspirin.json

Examples

We can pull cross-references for Aspirin (CID: 2244), saving the top 50 identifiers.

# Env: base-agent
python .agents/skills/general-chemical-literature/scripts/get_xrefs.py \
  --cid 2244 \
  --limit 50 \
  --outdir .agents/skills/general-chemical-literature/examples/aspirin \
  --output xrefs_aspirin.json

Constraints

• Novelty Assessment Limitation: If 0 PMIDs or Patents are returned, it strongly implies the molecule is highly novel (or purely computational), but it does not guarantee absolute non-existence.
• Link Generation: The script automatically prints actionable links (pubmed.ncbi.nlm.nih.gov/ and patents.google.com/patent/) for the top 5 results for immediate verification.
• Network Limits: Handled internally via standard exponential backoff.

---

---

Author: Bowen Deng Contact: GitHub @learningmatter-mit