GPURFSCREEN: a GPU based virtual screening tool using random forest classifier.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• In-silico methods are crucial for modern drug discovery.
• Virtual screening (VS) reduces the chemical space for experiments.
• VS requires substantial computation, making it time-consuming.

Purpose of the Study:

• To develop and evaluate a GPU-accelerated virtual screening tool.
• To speed up the virtual screening process for large bioassay datasets.

Main Methods:

• Utilized Random Forest, a robust classification algorithm.
• Implemented parallelized algorithms on Graphical Processing Units (GPUs).
• Developed the GPU-based Random Forest Virtual Screening (GPURFSCREEN) tool.

Main Results:

• GPURFSCREEN achieves optimized results with lower execution times on large datasets.
• The tool's performance on GPUs is comparable to serial environments in terms of result quality.
• Screened billions of molecules in training and prediction phases.

Conclusions:

• Parallelized virtual screening offers significantly reduced running times for high-throughput screening.
• The proposed parallel tool demonstrates superior performance over serial methods.
• GPU acceleration is effective for accelerating large-scale virtual screening in drug discovery.