Discovery of High-Affinity Amyloid Ligands Using a Ligand-Based Virtual Screening Pipeline

Area of Science:

• Neuroscience
• Medicinal Chemistry
• Computational Biology

Background:

• Fibrillar protein aggregates, such as amyloid-beta (Aβ(1-42)), are hallmarks of neurodegenerative diseases.
• Designing ligands for these aggregates is challenging due to limited structural information on binding sites.

Purpose of the Study:

• To develop and validate a computational, ligand-based virtual screening method for identifying novel, high-affinity ligands for Aβ(1-42) fibrils.
• To demonstrate the utility of this approach in discovering structurally unique amyloid ligands.

Main Methods:

• Assembled a database of known Aβ(1-42) ligands to train predictive models for dissociation constants based on chemical structure.
• Implemented a three-step virtual screening pipeline: molecular property filtering, machine learning using chemical descriptors, and 3D field point analysis.
• Screened 698 million compounds from the ZINC15 database and experimentally tested top candidates using thioflavin T fluorescence displacement assays.

Main Results:

• Identified five new Aβ(1-42) ligands with dissociation constants ranging from 20-600 nM.
• The novel ligands possess unique structures, validating the computational approach.
• Achieved an experimental hit rate of 10.9% from the virtual screening campaign.

Conclusions:

• The developed ligand-based virtual screening pipeline is effective for discovering novel, high-affinity amyloid ligands.
• This computational strategy offers a powerful tool for targeting difficult protein aggregate structures in neurodegenerative diseases.