An algorithm to classify homologous series within compound datasets.

Area of Science:

• Computational Chemistry
• Cheminformatics
• Data Science

Background:

• Homologous series are crucial in various scientific fields but often remain unannotated in chemical databases.
• This lack of annotation hinders the comprehension of chemical diversity and analytical identification.

Purpose of the Study:

• To develop and implement an algorithm for detecting homologous series within compound datasets.
• To enhance the annotation of chemical compounds and facilitate their identification.

Main Methods:

• Developed an algorithm using RDKit to classify homologous series based on repeating units (monomers) and core structures.
• Applied the algorithm to three diverse datasets: NORMAN-SLE (environmental chemistry), PubChemLite for Exposomics, and COCONUT (natural products).
• Utilized SMILES strings for molecules and SMARTS for monomers as primary inputs in an iterative process of substructure matching and fragmentation.

Main Results:

• Successfully classified over 2000, 12,000, and 5000 homologous series with CH2 repeating units in NORMAN-SLE, PubChemLite, and COCONUT datasets, respectively.
• Validated the algorithm's performance against published homologous series and existing methods, including PFAS compound categorization.
• Demonstrated the algorithm's effectiveness in identifying and grouping compounds belonging to homologous series.

Conclusions:

• The developed algorithm, OngLai, effectively classifies homologous series in large chemical datasets.
• This tool addresses a significant challenge in chemical data annotation, improving chemical diversity analysis and identification.
• The open availability of the algorithm promotes further research and application in related scientific domains.