SPRank─A Knowledge-Based Scoring Function for RNA-Ligand Pose Prediction and Virtual Screening.

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• RNA-targeted therapeutics are gaining importance, necessitating accurate computational models for RNA-small compound interactions.
• Existing scoring functions excel at predicting binding modes but lack thorough validation for virtual screening.
• Reliable scoring functions are crucial for both predicting native binding poses and ranking ligand affinities.

Purpose of the Study:

• To develop and validate SPRank, a novel scoring function for RNA-ligand interactions.
• To address the limitations of current scoring functions in virtual screening applications.
• To improve the accuracy of predicting both binding modes and affinities in RNA-ligand complex modeling.

Main Methods:

• SPRank integrates machine learning and knowledge-based approaches using a weighted iterative strategy.
• The method employs third-party docking software (rDock, AutoDock Vina) for flexible ligand sampling against RNA ensembles.
• Conformational flexibility of both RNA and ligands is captured to enhance interaction modeling.

Main Results:

• SPRank demonstrated superior performance over existing scoring functions across four diverse test sets (122, 42, 55, 71 complexes).
• The tool showed improved efficacy in virtual screening tasks, specifically for the HIV-1 TAR ensemble.
• Rigorous testing confirmed enhanced predictive capabilities for RNA-ligand complex interactions.

Conclusions:

• SPRank offers a significant advancement in computational modeling for RNA-ligand interactions.
• Its improved performance in binding mode prediction and virtual screening makes it a valuable tool for RNA-targeted drug design.
• The freely accessible source code and datasets facilitate further research and application in drug discovery.