xche-yds/agenticmedchem — explained in plain English
Analysis updated 2026-05-18
Extract chemical structures from PDF tables and convert drawings into SMILES text format.
Pull biological measurements like IC50 potency values out of scientific papers automatically.
Identify structure-activity relationship patterns across a set of drug candidate papers.
Process PROTAC and molecular glue papers by breaking bifunctional molecules into parts.
| xche-yds/agenticmedchem | alicankiraz1/codexqb | crain99/worldcut-2026 | |
|---|---|---|---|
| Stars | 28 | 28 | 28 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 4/5 | 3/5 | 3/5 |
| Audience | researcher | developer | general |
Figures from each repo's GitHub metadata at analysis time.
Requires a Claude CLI Max subscription, source PDFs and data are not included in the repo.
This repository is a Python pipeline that reads scientific papers and patents about drug development and extracts the useful data into a structured database. Drug researchers publish papers full of tables showing how different chemical compounds performed in lab tests, what their structures look like, how they were synthesized, and which proteins they target. Pulling all that information out by hand is slow and error-prone. This project automates the process using a team of AI agents that each handle a specific part of the job. The pipeline works in stages. First, it loads a PDF and renders its pages as images. Then several agents run at the same time: one extracts the chemical structures of compounds (expressed as SMILES, a text format that encodes molecular geometry), another pulls out the biological measurements like IC50 values (a standard way to express how potent a drug candidate is), a third reads synthesis routes, and a fourth identifies the protein targets. For dense tables of chemical structures that are drawn as images rather than text, the system uses optical recognition tools to convert those drawings into the SMILES format. After the parallel extraction phase, a reasoning agent identifies structure-activity relationships, meaning patterns in how small changes to a molecule's structure change its potency. A reviewer agent then re-reads the original PDF to check for errors or hallucinations in what the other agents produced. Results go into a SQLite database. The pipeline also handles a specialized category of drug molecules called PROTACs and molecular glues, which work by recruiting cellular machinery to destroy a disease-causing protein rather than just blocking it. Separate agents break these bifunctional molecules into their component parts. All AI calls go through the Claude CLI using a Max subscription rather than a pay-per-use API, so the cost is a flat monthly fee regardless of how many papers you process. The repository contains only code, source PDFs and extracted data are excluded from version control.
A multi-agent Python pipeline that extracts chemical structures and drug data from research papers into a database.
Mainly Python. The stack also includes Python, Claude CLI, SQLite.
Setup difficulty is rated moderate, with roughly 1h+ to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Verify against the repo before relying on details.