Neurogenerative diseases are often characterized by the formation of insoluble protein aggregates from amyloidogenic proteins. Alzheimer’s disease, currently the most common form of neurodegeneration, is specifically characterized by the deposition of amyloid plaques, comprised of misfolded amyloid- β (A-β) peptide and neurofibrillary triangles comprised of misfolded tau protein.

The only current way to definitively diagnose Alzheimer’s disease is through the post-mortem identification of A-β plaques. In the clinic, cognitive tests may be used, alongside a panel of imaging or biofluid tests, however these methods don’t reliably diagnose AD until extensive damage has already occurred.

Amyloid ligands are useful for ex vivo applications including for imaging and characterizing amyloid deposits or monitoring protein aggregation. To date, high-affinity amyloid ligand binders have been identified through high-throughput screening efforts, in combination with structure-activity-relationship (SAR) studies. However, the binding of amyloid ligands to fibrils is not a straightforward process. Multiple binding sites exist, and direct interactions between multiple ligands can occur.

Thanks to the large quantity of reported data for amyloid ligands, a ligand-based virtual screening method for finding high affinity ligands is appealing. This was the methodology used within a recent publication within the Journal of the American Chemical society, which I was pleased to co-author alongside Christopher Hunter and Timothy Chisholm at the University of Cambridge.

Within the publication, we presented a 3-step ligand based virtual screening approach, exploiting the wealth of A-β (1-42) ligand data available within scientific literature:

1. A set of 707 A-β (1-42) fibril-binding ligands were first combined, of which 388 held binding constants reported on the same binding site. Key molecular properties required for binding were identified from these, and a database of 698 million compounds were filtered using charge, molecular weight and logP, to prioritize 63 million compounds for further screening.
2. This subset was then screened by using two models developed by using the binding affinities of known amyloid ligands in a fused 6,5-benzoheterocycle (FBH) database.
3. These models were then used to select 100 compounds with the highest predicted binding affinity, 46 of which were experimentally investigated in fluorescence competition binding assays for A-β (1-42) fibrils.

The five highest affinity ligands all had nanomolar dissociation constants, representing a 10.9% hit rate for this virtual screening pipeline.

The publication is openly accessible via ACS Publications, and was published both online and in issue in July 2023.

Ligand based virtual screening at Cresset

Virtual screening is a valuable computational technique, which can deliver significant improvements for hit identification by increasing time efficiency and cost effectiveness during the initial stages of drug development, when compared to lab-based high-throughput screening methods.

Ligand based approaches require no prior knowledge of the biological target, as compounds are assessed based on their similarity to a reference compound which is known to be active. Ligand-based virtual screening platform, Blaze™ provides users with more diverse lead-like hits by using both the 3D electrostatics and shape of a defined ligand to rapidly search large chemical collections for ligands with similar properties.

Our team of Cresset Discovery CRO scientists also deliver outsourced virtual screening services on a contractual basis. With one of the most comprehensive virtual screening capabilities available in the computational chemistry outsourcing industry, we deliver both ligand based and structure based methods, tailored to your unique project requirements.