Vsevolod Vaskin, Dmitri Jakovlev, Fedor Bakharev

Substructure search in chemical compound databases is a fundamental task in cheminformatics with critical implications for fields such as drug discovery, materials science, and toxicology. However, the increasing size and complexity of chemical databases have rendered traditional search algorithms ineffective, exacerbating the need for scalable solutions. We introduce a novel approach to enhance the efficiency of substructure search, moving beyond the traditional full-enumeration methods. Our strategy employs a unique index structure: a tree that segments the molecular data set into clusters based on the presence or absence of certain features. This innovative indexing mechanism is inspired by the binary Ball-Tree concept and demonstrates superior performance over exhaustive search methods, leading to significant acceleration in the initial filtering process. Comparative analysis with the existing Bingo algorithm reveals the efficiency and versatility of our approach. Although the current implementation does not affect the verification stage, it has the potential to reduce false positive rates. Our method offers a promising avenue for future research, meeting the growing demand for fast and accurate substructure search in increasingly large chemical databases.