SAMPL4 & DOCK3.7: lessons for automated docking procedures.

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• The SAMPL4 challenge provided a benchmark for evaluating computational methods.
• Automated molecular docking pipelines are crucial for predicting molecular interactions.
• Accurate prediction of solvation energy and binding affinity remains a challenge in computational chemistry.

Purpose of the Study:

• To assess the performance of the DOCK 3.7 molecular docking pipeline.
• To evaluate automated methods for solvation energy, virtual screening, and pose/affinity prediction.
• To identify areas for improvement in molecular docking software.

Main Methods:

• Utilized the DOCK 3.7 pipeline within the SAMPL4 challenge framework.
• Tested automated methods for solvation energy, virtual screening, and binding affinity prediction.
• Analyzed performance against established benchmarks and other participating methods.

Main Results:

• DOCK 3.7's solvation energy predictions were five-fold worse than other SAMPL4 methods.
• Automated docking on the HIV Integrase allosteric site showed good virtual screening and pose prediction but poor affinity prediction.
• Default molecular docking grid sizes led to significant errors, necessitating adjustments for future use.

Conclusions:

• Lessons from SAMPL4 indicate a need for substantial improvements in molecular docking tools.
• Specific areas for enhancement include solvation energy and binding affinity prediction accuracy.
• Adjustments to parameters like grid size are critical for reliable molecular docking results.