blincs 0.1.0

Breadth-first Line Notation for Chemical Structures

BLINCS

Breadth-first Line Notation for Chemical Structures.

BLINCS is a compact line notation for molecules that encodes molecular structure as a breadth-first traversal of the spanning tree, recording ring closures as paired digits (similar to SMILES) but with a deterministic BFS ordering. It depends only on RDKit and the Python standard library.

A presentation from RDKit UGM 2025 is available here.

Installation

pip install blincs

Or from source:

git clone https://github.com/dehaenw/blincs.git
cd blincs
pip install .

Usage

Convert an RDKit mol to BLINCS:

from rdkit import Chem
from blincs import blincs

mol = Chem.MolFromSmiles("c1ccccc1")
bli = blincs.mol_to_blincs(mol)
# → 'cccccc%1%1'

Convert BLINCS back to an RDKit mol:

mol = blincs.blincs_to_mol("C1CCCC1")
# → RDKit RWMol for pentane

Use canonical=True for stereo-aware canonical encoding (recommended):

bli = blincs.mol_to_blincs(mol, canonical=True)

More examples in examples.ipynb.

Roundtrip accuracy

Benchmarked on ChEMBL 34 (2,310,851 molecules) — full InChI comparison:

Category	Count	%
Full match (identical InChI)	2,310,815	99.998%
Parse failures (RDKit cannot parse input)	4	0.0002%
Decode failures (exotic polycyclic cages)	7	0.0003%
Stereo mismatches	25	0.001%

The 25 remaining stereo mismatches are concentrated in highly symmetric cage systems (bispyridinium cages, adamantane/norbornane bridges, and symmetric halocyclohexanes) where CIP-based parity assignment is ambiguous.

On the 2,000-molecule stereo-containing subset (data/chembl_2k.smi), roundtrip accuracy is 100% (all stereochemistry preserved).

Testing

pip install pytest
pytest tests/

License

MIT — see LICENSE.