Redesigning Vina@QNLM for Ultra-Large-Scale Molecular Docking and Screening on a Sunway Supercomputer.

Area of Science:

• Computational chemistry
• Drug discovery informatics
• High-performance computing

Background:

• Accurate identification of lead compounds is crucial for efficient drug discovery.
• Ultra-large-scale molecular docking offers enhanced precision in identifying potential drug candidates.
• Existing computational methods face limitations in handling massive compound libraries.

Purpose of the Study:

• To develop and implement a novel molecular docking software, Vina@QNLM, capable of ultra-large-scale parallel processing.
• To optimize molecular docking algorithms for heterogeneous multicore processor architectures.
• To establish a new benchmark for parallel computing in molecular docking applications.

Main Methods:

• Development of Vina@QNLM software leveraging a task scheduling mechanism for large-scale parallelism on Sunway supercomputers.
• Adaptation of the core docking algorithm to exploit heterogeneous multicore processor architectures for intensive computing.
• Expansion of computational resources to over 10 million cores, achieving a scalability of 55.92%.

Main Results:

• Vina@QNLM successfully scaled to 10,465,065 cores, marking the first use of 10 million cores for molecular docking on Sunway architecture.
• Heterogeneous multicore processor integration yielded an 11x speedup compared to management process elements.
• Vina@QNLM demonstrated superior screening power, ranking highest among 27 software packages in the CASF-2013 benchmark.

Conclusions:

• Vina@QNLM represents a significant advancement in ultra-large-scale molecular docking, enabling unprecedented computational power for drug discovery.
• The software's high performance and scalability facilitate rapid screening of vast compound libraries against therapeutic targets.
• A publicly accessible platform for Vina@QNLM was developed, democratizing access to advanced molecular docking capabilities.