SAFR: Enabling Fragment-Based Drug Discovery with a Synthetic Binding Pose Data Set

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Fragment-Based Drug Discovery (FBDD) is effective for identifying potent drug candidates from small fragments.
• Computational FBDD relies heavily on high-quality structural data of fragment binding poses, which is often limited.
• Existing databases may lack sufficient coverage of diverse fragment-protein interactions.

Purpose of the Study:

• To introduce the Structurally Augmented Fragment Repository (SAFR), a novel dataset designed to address data limitations in computational FBDD.
• To provide a comprehensive resource of fragment-protein interactions to support in silico drug design.
• To facilitate fragment screening and hit-to-lead optimization in drug discovery campaigns.

Main Methods:

• A filtering protocol was applied to 89,375 high-confidence binding poses from public sources, using 2D similarity, 3D superposition, docking, and interaction features.
• Bioactive ligands were fragmented within their predicted binding poses to generate fragment-protein interactions.
• The resulting dataset was analyzed for novelty compared to existing protein-bound ligand structures in the Protein Data Bank (PDB).

Main Results:

• The SAFR dataset contains 818,385 fragment-protein interactions involving 157,080 unique fragments and 1,142 proteins.
• 270,155 unique fragment-protein interactions were identified.
• A significant majority (88%) of these unique interactions (237,284) are not represented in existing PDB structures, highlighting the novelty of the dataset.

Conclusions:

• SAFR is a valuable new resource for computational Fragment-Based Drug Discovery.
• The dataset effectively supports in silico fragment screening and hit-to-lead optimization.
• SAFR provides novel fragment-protein interaction data, expanding the possibilities for drug design.