Get Your Atoms in Order--An Open-Source Implementation of a Novel and Robust Molecular Canonicalization Algorithm

Area of Science:

• Cheminformatics
• Computational Chemistry
• Drug Discovery

Background:

• Canonical molecular ordering is crucial for unique molecule representation.
• Existing methods like the Morgan algorithm have limitations with large molecules and inconsistencies across toolkits.
• Lack of a universal standard hinders molecule identification.

Purpose of the Study:

• To develop an alternative canonicalization approach for molecules.
• To address limitations of current graph relaxation algorithms.
• To facilitate the generation of universal unique molecule identifiers.

Main Methods:

• Utilized a standard stable-sorting algorithm instead of Morgan-like indexing.
• Developed two new invariants for handling dependent chirality and symmetrical cyclic graphs.
• Tested the algorithm on the ChEMBL 20 dataset (1.45 million compounds).

Main Results:

• The new approach demonstrated robustness and speed, even with random atom renumbering and SMILES round tripping.
• Achieved canonical ordering for protein molecules in milliseconds.
• The algorithm is implemented in the RDKit toolkit and provided with a Python reference implementation.

Conclusions:

• The novel algorithm provides a reliable and efficient method for canonical atom ordering.
• This work represents a significant step towards a common standard for universal molecule identification.
• The open-source implementation facilitates integration into various cheminformatics toolkits.