Searching for Recursively Defined Generic Chemical Patterns in Nonenumerated Fragment Spaces

Retrieving molecules with specific structural features is a fundamental requirement of today's molecular database technologies. Estimates claim the chemical space relevant for drug discovery to be around 10⁶⁰ molecules. This figure is many orders of magnitude larger than the amount of molecules conventional databases retain today and will store in the future. An elegant description of such a large chemical space is provided by the concept of fragment spaces. A fragment space comprises fragments that are molecules with open valences and describes rules how to connect these fragments to products. Due to the combinatorial nature of fragment spaces, a complete enumeration of its products is intractable. We present an algorithm to search fragment spaces for generic chemical patterns as present in the SMARTS chemical pattern language. Our method allows specification of the chemical surrounding of an atom in a query and, therefore, enables a chemically intuitive search. During the search, the costly enumeration of products is avoided. The result is a fragment space that exactly describes all possible molecules that contain the user-defined pattern. We evaluated the algorithm in three different drug development use-cases and performed a large scale statistical analysis with 738 SMARTS patterns on three public available fragment spaces. Our results show the ability of the algorithm to explore the chemical space around known active molecules, to analyze fragment spaces for the presence of likely toxic molecules, and to identify complex macromolecular structures under additional structural constraints. By searching the fragment space in its nonenumerated form, spaces covering up to 10¹⁹ molecules can be examined in times ranging between 47 s and 19 min depending on the complexity of the query pattern.