Computational identification of proteins for selectivity assays.

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Drug optimization requires assessing compound binding to target proteins and a selectivity panel.
• Current methods for selecting selectivity panel proteins rely heavily on sequence homology, potentially missing nonhomologous targets.
• Experimental selectivity data is often limited or unavailable during early drug development.

Purpose of the Study:

• To develop a computational method for identifying potential off-target proteins for selectivity panels.
• To overcome limitations of homology-based selection by focusing on binding site similarity.
• To enable the construction of more comprehensive selectivity panels early in drug discovery.

Main Methods:

• Developed a computational approach using docking scores of target-selected molecular probes to evaluate binding site similarity.
• Generated molecular probes by docking a diverse library of drug-like compounds to the target protein.
• Applied the method to proteins with known 3D structures, independent of sequence homology or known inhibitor data.

Main Results:

• The method effectively identifies proteins with similar binding sites to the target, even those lacking sequence homology.
• Successfully rediscovered known nonhomologous protein binders for common ligands like estradiol, tamoxifen, and riboflavin.
• Demonstrated the ability to discriminate proteins with similar binding sites from random proteins based on 3D structure.

Conclusions:

• The developed computational method provides an effective strategy for identifying selectivity panel proteins.
• This approach enhances the identification of potential off-target interactions early in the drug discovery process.
• The method is broadly applicable to any protein with a known 3D structure, facilitating more robust drug development.