1001 Ways to run AutoDock Vina for virtual screening

Area of Science:

• Computational chemistry and drug discovery.
• High-performance computing (HPC) applications in life sciences.

Background:

• Large-scale virtual screening requires significant computational resources.
• Choosing appropriate high-performance computing (HPC) infrastructure is challenging for biochemists.
• Understanding technical alternatives impacts the efficiency of virtual screening experiments.

Purpose of the Study:

• To review considerations for running large virtual screening experiments with AutoDock Vina.
• To guide scientists in selecting optimal computing platforms and configurations.
• To illustrate the impact of parallelization, reproducibility, and execution time factors on screening throughput.

Main Methods:

• Analysis of AutoDock Vina performance on different HPC infrastructures (grid, Hadoop, small cluster, multi-core virtual machine).
• Experimental evaluation of parallelization strategies and reproducibility measures.
• Investigation of factors influencing per-ligand execution time (active torsions, heavy atoms, exhaustiveness).

Main Results:

• Increased parallelization enhances virtual screening throughput on multi-core systems.
• Random seed capture is insufficient for reproducibility on heterogeneous distributed systems.
• Optimizing factors like active torsions and exhaustiveness improves overall screening time.
• Different HPC platforms suit virtual screening experiments of varying scales.

Conclusions:

• Scientists can improve virtual screening efficiency by understanding computational factors and platform suitability.
• Informed choices regarding HPC infrastructure and AutoDock Vina configuration are crucial for large-scale drug discovery.
• This review provides practical guidance for optimizing future virtual screening endeavors.