Rosetta's Predictive Ability for Low-Affinity Ligand Binding in Fragment-Based Drug Discovery

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Fragment-based drug discovery (FBDD) identifies weak-binding small molecules for drug development.
• Computational docking in FBDD faces challenges due to limited binding interactions and sensitivity to energy functions/poses.
• Accurate computational screening is crucial for ranking potential drug fragments.

Purpose of the Study:

• To evaluate RosettaLigand's performance in fragment-based drug discovery.
• To compare in silico docking results with experimental NMR spectroscopy data.
• To assess RosettaLigand's ability to rank fragments and identify binding poses.

Main Methods:

• Docking of 3456 fragments to HisF (a cysteine-depleted TIM-barrel protein) using RosettaLigand.
• Experimental screening of fragments using NMR spectroscopy to determine binding affinities and locations.
• Comparison of computational predictions with experimental NMR data for ranking and pose validation.

Main Results:

• Identified 31 high-affinity binders from 3456 fragments, with dissociation constants as low as 400 μM.
• RosettaLigand achieved an Area Under the Receiver Operating Characteristic Curve (AUC-ROC) of 0.74 in ranking binders vs. non-binders.
• Docking poses predicted by RosettaLigand for binders accurately matched the experimentally determined binding pocket.

Conclusions:

• RosettaLigand demonstrates reliable performance in fragment-based drug discovery settings.
• The study provides a benchmark for RosettaLigand's capabilities in virtual screening and pose prediction.
• Integration of computational and experimental methods enhances FBDD efficiency.