Skillget

pubchem-finder

mcp
Security Audit
Warn
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 6 GitHub stars
Code Pass
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Molecular similarity search over PubChem with RDKit: CLI, MCP server for Claude, local fingerprint index

README.md

PubChem Finder

Molecular similarity search over PubChem, built on RDKit.
The goal is a tool you can run on a laptop: build a local fingerprint index of PubChem
Compound, search it from the command line, and expose it to Claude (or any MCP client)
as an MCP server.

CI
Python 3.11+
License: MIT

Status: v0.1 of the rebuild. PubChem lookups, RDKit descriptors,
standardization and fingerprint comparison work from the CLI and from an MCP
server. The local similarity index over all of PubChem is the next milestone.
See ROADMAP.md.

Quick start

Requires uv. No install step: uvx fetches and caches the tool.

# Look a compound up in PubChem (by name, CID, SMILES or InChIKey)
uvx --from git+https://github.com/MysterionRise/pubchem-finder pubchem-finder lookup caffeine
uvx --from git+https://github.com/MysterionRise/pubchem-finder pubchem-finder lookup 2244 --by cid --json

# RDKit descriptors, standardization, similarity between two structures (offline)
uvx --from git+https://github.com/MysterionRise/pubchem-finder pubchem-finder props "CC(=O)Oc1ccccc1C(=O)O"
uvx --from git+https://github.com/MysterionRise/pubchem-finder pubchem-finder standardize "CC(=O)[O-].[Na+]"
uvx --from git+https://github.com/MysterionRise/pubchem-finder pubchem-finder compare "CCO" "CCCO"

Or install once: uv tool install git+https://github.com/MysterionRise/pubchem-finder, then use pubchem-finder ....

Use it from Claude (MCP)

The same operations are available as an MCP server over stdio.

Claude Code:

claude mcp add pubchem-finder -- uvx --from git+https://github.com/MysterionRise/pubchem-finder pubchem-finder mcp

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "pubchem-finder": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/MysterionRise/pubchem-finder", "pubchem-finder", "mcp"]
    }
  }
}

Tools exposed:

Tool What it does
pubchem_lookup(query, by, limit) Resolve a name, CID, SMILES or InChIKey to PubChem records (identifiers, formula, MW, XLogP, H-bond counts, TPSA, link). Live PubChem PUG REST.
compound_properties(smiles) RDKit descriptors and Lipinski check for any SMILES. Offline.
standardize(smiles) Remove salts and counter-ions, neutralize, canonicalize. Offline.
compare(smiles_a, smiles_b) Tanimoto similarity on ECFP4 and MACCS fingerprints. Offline.

Then ask Claude things like "what is the molecular weight and XLogP of caffeine?" or
"how similar are aspirin and salicylic acid?".

Library

The CLI and MCP server are thin layers over pubchem_finder.service; the chemistry
lives in pubchem_finder.chemistry and pubchem_finder.fingerprints.

from pubchem_finder import service

service.properties("CC(=O)Oc1ccccc1C(=O)O").molecular_weight  # 180.16
service.standardize("CC(=O)[O-].[Na+]").smiles  # 'CC(=O)O'
service.compare("CCO", "CCCO").ecfp4_tanimoto  # 0.556
service.lookup("caffeine").compounds[0].cid  # 2519 (network)

Lower level:

from pubchem_finder.chemistry import MoleculeStandardizer, calculate_properties
from pubchem_finder.fingerprints import MorganFingerprint, TanimotoSimilarity

gen = MorganFingerprint(radius=2, n_bits=1024)  # ECFP4, 1024 bits
aspirin = gen.generate_from_smiles("CC(=O)Oc1ccccc1C(=O)O")
salicylic = gen.generate_from_smiles("O=C(O)c1ccccc1O")
print(TanimotoSimilarity().calculate(aspirin, salicylic))

Development

Requires uv. Python 3.11, 3.12 and 3.13 are supported.

uv sync            # create .venv and install everything, including dev tools
make check         # ruff lint + format check, mypy (strict), pytest with coverage
make test          # just the tests
PUBCHEM_LIVE=1 uv run pytest -m live   # the one test that calls the real PubChem API

Pre-commit hooks (ruff lint and format) can be installed with uv run pre-commit install
after uv tool install pre-commit.

Planned architecture

PubChem Compound SDF (HTTPS) --> streaming RDKit parse --> Parquet metadata (CID, SMILES, InChIKey, MW, formula)
                                                      \-> FPSim2 fingerprint index (ECFP4, 1024 bits)
pubchem-finder CLI  ----\
                         +--> search facade --> FPSim2 (exact Tanimoto, in-memory or on-disk) + DuckDB
MCP server (stdio) -----/                   \-> PubChem PUG REST for live lookups without an index

The index engine is FPSim2, the similarity
search library behind ChEMBL's interface. A PostgreSQL + RDKit cartridge backend for
substructure search is an optional later addition behind the same interface.

History

This repository started in 2020 as an Indigo + Elasticsearch indexer for PubChem
dumps. That code has been removed; it is available in the git history before
commit 06619d9 if you need it.

License

MIT. See LICENSE.

Reviews (0)

No results found