checking-chembl_skill

Overview

This skill checks whether a medicinal chemistry paper is present in the ChEMBL database so you can access curated, machine‑readable bioactivity data instead of parsing PDFs. It prioritizes DOIs for lookups and returns ChEMBL document IDs and activity counts to inform whether structured SAR data is available. Use the result to skip manual extraction when high‑quality curated data exists.

How this skill works

The skill queries the ChEMBL document endpoint by DOI only (ChEMBL does not accept PMID queries). If a document is found it extracts the document_chembl_id, then queries the activity endpoint to obtain the total number of bioactivity points. Results include doc_type, title, ChEMBL ID, and activity count so you can decide whether to pull full activity JSON or continue with PDF parsing.

When to use it

• After abstract screening with relevance score ≥7
• When a paper reports IC50, Ki, EC50, MIC or other activity measurements
• Before attempting table or figure parsing of PDFs
• When you want standardized SMILES, units, and validated values
• When preparing summaries or updating a papers-reviewed tracking file

Best practices

• Always query ChEMBL by DOI; fetch DOI from PubMed if you only have a PMID
• Report ChEMBL status immediately in your workflow and add the ChEMBL ID to summaries
• If activity_count > ~20 consider extracting full activity JSON for analysis
• Still fetch full text for methods, synthesis, and interpretation even if ChEMBL data exists
• Record ChEMBL status and activity_count in your papers-reviewed.json or summary file

Example use cases

• Screening a batch of medicinal chemistry hits: skip PDF extraction for papers found in ChEMBL
• Adding ChEMBL IDs and activity counts to SUMMARY.md during a literature review
• Automated scripts that flag papers with >50 curated data points for deeper SAR analysis
• Pulling all IC50 values for a target from ChEMBL for quick potency statistics
• Deciding whether to trust curated data or parse original figures when preparing datasets