Computing Ligands Bound to Proteins Using MELD-Accelerated MD.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Virtual screening methods like DOCK are crucial for predicting small-molecule ligand poses in protein binding sites.
• Atomistic molecular dynamics (MD) offers superior free-energy-based pose selection but is computationally intensive.
• Existing methods often struggle with accurate pose prediction for complex ligand-protein interactions.

Purpose of the Study:

• To evaluate the efficacy of MELD-accelerated MD (MELD × MD) in identifying correct ligand poses from DOCK outputs.
• To determine if MELD × MD can improve upon the limitations of traditional virtual screening and standard MD simulations.
• To assess the added value of MELD × MD for accurate small-molecule pose prediction.

Main Methods:

• Utilized MELD-accelerated MD (MELD × MD) simulations on 30 distinct ligand-protein pairs.
• Input poses were generated using the DOCK virtual screening algorithm.
• Evaluated pose accuracy based on free energy calculations and identification of native poses.

Main Results:

• MELD × MD successfully identified native ligand poses in 23 out of 30 studied cases.
• The method corrected inaccurate poses from DOCK in 20 of these successful predictions.
• Free energy calculations correlated with the identification of the most accurate ligand poses.

Conclusions:

• MELD × MD significantly enhances the accuracy of small-molecule pose prediction in protein binding sites.
• This accelerated MD approach offers a valuable improvement over standard virtual screening techniques.
• MELD × MD demonstrates potential for improving drug discovery pipelines by refining ligand pose selection.