Searching Fragment Spaces with feature trees.

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Drug Discovery

Background:

• Combinatorial chemistry generates vast numbers of virtual compounds.
• Conventional virtual screening methods are inadequate for such large chemical spaces.
• Fragment Spaces offer a solution by encoding compounds via fragments and synthesis rules.

Purpose of the Study:

• To describe the generation of a large Fragment Space encoding approximately 5 * 10^11 compounds.
• To demonstrate the integration of Fragment Space searches into an overall virtual screening workflow.
• To reduce the number of virtual compounds for more manageable downstream analysis.

Main Methods:

• Encoding established in-house synthesis protocols for combinatorial libraries into a machine-readable format.
• Developing similarity-based searches within the generated Fragment Space.
• Integrating Fragment Space searches as a preliminary step in a larger in silico workflow.

Main Results:

• Successfully generated a Fragment Space representing about 5 * 10^11 virtual compounds.
• Demonstrated that Fragment Space searches can reduce the number of candidate molecules by several orders of magnitude.
• Case study results show the practical utility of this approach in pharmaceutical research.

Conclusions:

• Fragment Spaces provide an efficient method for searching extremely large virtual chemical libraries.
• This approach enables in silico screening of vast compound collections, accelerating drug discovery.
• The methodology offers a scalable solution for expanding the chemical space accessible to pharmaceutical companies.