ENRI: A tool for selecting structure-based virtual screening target conformations.

Area of Science:

• Computational chemistry
• Structural biology
• Machine learning in drug discovery

Background:

• Structure-based virtual screening (SBVS) faces challenges in identifying pharmaceutically relevant protein conformations.
• The increasing number of experimentally determined and computationally derived protein conformations complicates target selection.
• Conformational complexity in drug targets poses a significant hurdle for virtual screening efficacy.

Purpose of the Study:

• To develop a novel machine learning method for selecting druggable receptor conformations.
• To address the challenge of conformational complexity in structure-based virtual screening.
• To identify pharmaceutically interesting receptor conformations for nuclear receptors and other drug targets.

Main Methods:

• Utilized machine learning, specifically an over-sampling and binary classification procedure.
• Trained a binary classifier on nuclear receptor conformations labeled with structure-based virtual screening enrichment measures.
• Developed a methodology applicable to nuclear receptors and extendable to other protein targets.

Main Results:

• Successfully identified enriching structure-based virtual screening targets for six out of seven nuclear receptors.
• Demonstrated the ability to formulate suggestions for selecting druggable receptor conformations.
• Validated the classifier's potential for extension to diverse protein targets with new training data.

Conclusions:

• The proposed machine learning method effectively selects pharmaceutically relevant receptor conformations.
• This approach enhances the efficiency of structure-based virtual screening by navigating conformational complexity.
• The methodology offers a scalable solution for identifying viable drug targets across various proteins.