Protein-ligand interaction prediction: an improved chemogenomics approach

Area of Science:

• Computational chemistry and cheminformatics
• Drug discovery and development
• Structural biology and bioinformatics

Background:

• Predicting small molecule-protein interactions is vital for understanding biological processes and advancing drug discovery.
• Ligand-based virtual screening models struggle with accuracy when few known ligands are available, especially for orphan receptors.
• The absence of 3D protein structures limits traditional structure-based drug design approaches.

Purpose of the Study:

• To develop a systematic method for predicting ligand-protein interactions applicable to targets with no 3D structure and few or no known ligands.
• To improve the accuracy of predictive models for drug targets with limited available data.
• To address the limitations of classical ligand-based virtual screening in challenging drug discovery scenarios.

Main Methods:

• Implemented a chemogenomics strategy by screening chemical space against entire protein families simultaneously.
• Utilized known ligands from similar protein targets to compensate for the lack of data for specific targets.
• Tested the cross-target prediction strategy on enzymes, G-protein-coupled receptors (GPCRs), and ion channels.

Main Results:

• Achieved significant improvements in prediction accuracy compared to classical ligand-based virtual screening.
• Demonstrated enhanced performance particularly for targets with few or no known ligands.
• Successfully predicted ligand-protein interactions for challenging targets across diverse drug target classes.

Conclusions:

• The proposed cross-target chemogenomics approach effectively predicts ligand-protein interactions, overcoming data limitations.
• This method offers a powerful alternative for drug discovery, especially for orphan receptors and targets with sparse data.
• The strategy significantly advances the capability to screen chemical libraries against a broad range of protein targets.