Shape-based methods for Structure-based Fragment Growing

Abstract

Fragment-based drug design is an influential paradigm in pharmaceutical development. It has revolutionized experimental approaches and methodological concepts. However, computational support of fragment-based drug design often consists of software made for full-sized ligands. This generalization fails to address the differences between ligands and fragments. Furthermore, the validation of these methods is often not performed with statistically significant sample sizes. In this project, we developed a new tool for structure-based fragment growing that addresses these issues.

The tool was developed from the ground up by starting with the fundamental question of torsion angle preferences in small molecules. A long-standing torsion library system was regenerated with new data and large parts of the infrastructure reworked. In the course of this work, significant methodological improvements were made to the system using recently developed SMARTS technology, which led to chemically meaningful improvements in the torsion angle statistics.

On top of this we built the core of our tool, which is a novel shape-based algorithm that exploits the directionality of fragment growing. Shape descriptors could be generated symmetrically for pockets and fragments, and could be compared at high speeds. The quality of the results was measured using a set of fragment growing steps, which were extracted from public crystal structure information.

A full-fledged modeling tool was built around the shape-based algorithm. This required a number of other components, such as geometry optimization. The workflow also included a pharmacophoric constraints system to search for fragments that generate specific interactions and an ensemble flexibility implementation to handle multiple conformations of a binding site. The tool itself was compared to an established structure-based method. Furthermore, the tool was applied in a case study to demonstrate a practical application. Our efforts have continually been close to active drug development projects and the tool is now in use at several organizations.