DrugScoreRNA--knowledge-based scoring function to predict RNA-ligand interactions.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• RNA's growing importance as a drug target necessitates advanced structure-based drug design tools.
• Current docking and scoring methods require improvement for accurate prediction of RNA-ligand interactions.

Purpose of the Study:

• To develop a novel knowledge-based scoring function, DrugScoreRNA, for predicting RNA-ligand interactions.
• To evaluate DrugScoreRNA's performance in predicting binding modes and affinities compared to existing tools.

Main Methods:

• Derived distance-dependent pair potentials from 670 crystallographically determined nucleic acid-ligand and -protein complexes.
• Applied DrugScoreRNA as an objective function for docking 31 RNA-ligand complexes and 20 NMR structures.
• Predicted binding affinities for 15 RNA-ligand complexes and compared with experimental data.

Main Results:

• DrugScoreRNA achieved "good" binding geometries (RMSD < 2 Å) in 42% of cases on the first rank, outperforming other methods.
• Demonstrated robustness with good docking results for NMR structures not included in the knowledge base.
• Showed a fair correlation (R S = 0.61) between computed and experimental binding affinities, sufficient for virtual screening.

Conclusions:

• DrugScoreRNA is a robust and efficient tool for RNA-ligand docking and binding affinity prediction.
• The developed potentials facilitate faster convergence to the global minimum during docking.
• DrugScoreRNA offers superior predictive power, advancing structure-based drug design for RNA targets.