Visually interpretable models of kinase selectivity related features derived from field-based proteochemometrics

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Achieving selectivity for small organic molecules toward biological targets like kinases is challenging due to similar ATP binding pockets.
• This difficulty leads to side effects and limits therapeutic efficacy.
• There is a need for improved methods to design selective inhibitors.

Purpose of the Study:

• To develop a novel computational method combining ligand- and receptor-based information for designing selective kinase inhibitors.
• To support the creation of drugs with fewer side effects or altered target profiles for enhanced efficacy.
• To enable better understanding of interactions between multiple ligands and multiple proteins.

Main Methods:

• Utilized proteochemometrics, a multivariate statistics approach, to correlate ligand and protein descriptions with receptor affinity.
• Described superimposed binding sites of 50 unique kinases using molecular interaction fields (MIFs) from knowledge-based potentials and WaterMap.
• Characterized 80 ligands using Mold(2), Open Babel, and Volsurf descriptors.
• Employed partial least-squares regression with cross-terms for model building to describe selectivity.

Main Results:

• Developed a predictive model combining ligand and kinase structural information.
• The models allowed for interpretation and visualization within the context of ligand binding pockets.
• Demonstrated a method to correlate molecular descriptors with binding affinity across multiple targets.

Conclusions:

• The developed proteochemometrics method effectively integrates ligand and protein data for drug design.
• This approach facilitates the design of novel, selective kinase inhibitors.
• The method offers a visualizable and interpretable framework for optimizing drug selectivity and efficacy.