Iterative Knowledge-Based Scoring Functions Derived from Rigid and Flexible Decoy Structures: Evaluation with the...

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Accurate prediction of protein-ligand interactions is crucial for drug discovery.
• Existing scoring functions have limitations in predicting binding modes and affinities.
• Knowledge-based scoring functions offer a promising approach for improving prediction accuracy.

Purpose of the Study:

• To develop and evaluate novel iterative knowledge-based scoring functions for protein-ligand complexes.
• To assess the performance of these new functions against established methods using benchmark datasets.
• To investigate the potential for developing protein-family-dependent scoring functions.

Main Methods:

• Generation of rigid and flexible decoy structures from PDBbind 2012 protein-ligand complexes.
• Development of two iterative knowledge-based scoring functions: ITScore_pdbbind(rigid) and ITScore_pdbbind(flex).
• Evaluation using the 2013 and 2014 CSAR benchmarks and comparison with Vina and ITScore.
• Development of a graph-based method for evaluating conformational root-mean-square deviation.

Main Results:

• The new ITScore_pdbbind scoring functions demonstrated significantly improved binding mode prediction performance.
• All evaluated scoring functions exhibited protein-dependent performance for binding affinity predictions.
• A novel graph-based method for comparing ligand conformations was developed and made freely available.

Conclusions:

• Iterative knowledge-based scoring functions trained on larger datasets enhance binding mode prediction accuracy.
• Protein-specific or protein-family-specific models are likely necessary for accurate binding affinity predictions.
• The developed tools and methods contribute to advancing computational drug design and molecular modeling.