Facing the challenges of structure-based target prediction by inverse virtual screening

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Structure-based computational target prediction is crucial for assessing off-target effects and understanding drug mechanisms.
• Existing methods face challenges in accuracy and efficiency, necessitating improved approaches.

Purpose of the Study:

• To introduce iRAISE, a novel inverse screening method for efficient and accurate computational target prediction.
• To address limitations of sequential pairwise docking and scoring noise in protein-ligand interactions.

Main Methods:

• Developed iRAISE, an inverse screening engine utilizing triangle descriptors.
• Implemented a Scoring Cascade incorporating reference ligand, ligand, and active site coverage.
• Applied statistical evaluation of score cutoffs for individual protein pockets.

Main Results:

• iRAISE achieved high accuracy, ranking over 35% of targets first and predicting >80% of binding modes within 2.0 Å RMSD on the Astex Diverse Set.
• Demonstrated rapid screening with a median computation time of 5 seconds per protein.
• On a large dataset, iRAISE identified the first true positive within the top 8 ranks (median).

Conclusions:

• iRAISE offers a significant advancement in structure-based computational target prediction.
• The method enables rapid screening of large protein structure datasets for drug discovery.
• iRAISE effectively predicts both target identity and binding modes, aiding in rational drug design.