UNSUPERVISED NAMING OF CHEMICAL COMPOUNDS

Miglė Urbonaitė1, Andrius Merkys2

1 Department of Mathematical Computer Science, Vilnius University, Lithuania

2 Institute of Biotechnology, Life Sciences Center, Vilnius University, Lithuania

The chemical diversity that already exists in the world is surprisingly extensive, and over time it is only expanding. Today, according to the CAS registry, almost 200 million chemicals and chemical compounds are registered for production and use. The natural problem that occurs is the identification of all these chemical compounds. Without unique names, the ability to unambiguously describe chemicals in literature, databases, and spoken language is lost.

However, identification by numbers or other index-based synthetic codes, although always unique, is not always convenient because without the index it says nothing about the compound or its properties [1]. In the chemical world, this problem is solved by nomenclatures. They describe the rules by which molecules are unambiguously named and the original molecules can be unambiguously reconstructed from these names [2]. One of the most popular and widely used is the IUPAC nomenclature. When using the nomenclature, it is very important that the name formed is unique and has an unambiguous relationship with the compound, as the chemical compounds may be very similar but have completely opposite properties (a well-known example is the stereochemistry of drug Thalidomide that caused birth defects). Thus, it is very important for the name to reflect each and every property of the compound. This is one of the research subjects of the field of cheminformatics [3]. Using chemical molecular graphs and graph theory algorithms, software packages, which can cover a large set of nomenclature rules and create unambiguous and correct names for chemical compounds, are created.

Here we report the development of ChemOnomatopist [4], a free and open source software package to name chemical compounds according to the IUPAC nomenclature. Our method employs graph algorithms to implement IUPAC nomenclature rules, as opposed to database or machine-learned model-based tools like STOUT [5]. ChemOnomatopist is able to read SMILES linear notation and output IUPAC names without supervision, opening a possibility to process large volumes of chemical data.


[1] J.-L. Faulon, D. P. Visco Jr and D. Roe, Enumerating molecules. Reviews in Computational Chemistry 21, 209-286 (2005).

[2] A. J. Williams and A. Yerin, Automated systematic nomenclature generation for organic compounds, Wiley Interdisciplinary Reviews: Computational Molecular Science 3.2, 150-160 (2013).

[3] T. Engel, Basic overview of chemoinformatics, Journal of chemical information and modeling 46.6, 2267-2277 (2006).

[4] A. Merkys. https://github.com/merkys/ChemOnomatopist

[5] K. Rajan, A. Zielesny and C. Steinbeck, STOUT: SMILES to IUPAC names using neural machine translation, Journal of Cheminformatics 13, 34 (2021). https://doi.org/10.1186/s13321-021-00512-4