QueryPubChemVendors

Reaxys → PubChem enrichment (CID, SMILES, vendor availability)

This repository contains a Jupyter notebook workflow that takes a Reaxys search results CSV, cleans/deduplicates InChI Keys, and enriches each unique compound using PubChem with:

• CID (PubChem compound identifier)
• SMILES (single SMILES field)
• Vendor availability (PubChem “Chemical Vendors” source category)

The notebook is written to be practical on larger datasets by deduplicating keys up front, using retries/backoff for PubChem requests, and caching results so reruns are much faster.

---

Files in this repo

Typical layout (adjust to match your repo):

• notebooks/query_pubchem.ipynb — main workflow
• README.md
• environment.yml — dependencies
• .gitignore — ignore cache files / large inputs

---

Input: Reaxys CSV requirements

Your Reaxys export CSV must contain these columns (exact names):

• Reaxys Registry Number
• CAS Registry Number
• Chemical Name
• InChI Key

The workflow will:

• strip whitespace from InChI Key
• drop empty / missing keys
• (optionally) create a row-deduplicated table with one row per unique key

---

Output

The final table contains the original Reaxys columns plus:

• CID
• SMILES
• VendorCount — number of unique PubChem vendor source names
• Vendors — semicolon-separated preview of vendor names (first N; configurable)

---

Quickstart

1) Clone this repo

git clone https://github.com/saramccormack/QueryPubChemVendors.git
cd QueryPubChemVendors

2) Create an environment

Using conda (recommended)
conda env create -f environment.yml
conda activate <env-name>

3) Launch Jupyter and run the notebook

jupyter lab

Open notebooks/query_pubchem.ipynb and run cells top-to-bottom.

Performance notes (important for big datasets)

This workflow is network-bound (PubChem API calls). It is faster and more reliable when you:

Deduplicate InChI Keys before querying (major speedup if the CSV has repeats)

Use retry/backoff to handle transient connection drops or rate limiting

Use caching so reruns don’t re-query keys that already succeeded

If you run in parallel (threaded), typical tuning knobs are:

MAX_WORKERS: reduce if you see disconnects; increase cautiously if stable

request throttling/sleep: increase if you see many retries

How vendor availability is determined

Vendor information is pulled from PubChem PUG-View categories for each CID and filtered to the category:

Category == "Chemical Vendors"

Vendor “labels” are taken from SourceName entries in that category.

Note: PubChem vendor listings can change over time and the ordering is not guaranteed. This workflow reports a count and an optional preview list for convenience.

License

MIT License © 2026 Sara McCormack. See the LICENSE file for details.