PyRMD Studio: A Unified Suite for Next-Generation, AI-Powered Virtual Screening

Area of Science:

• Computational chemistry
• cheminformatics
• drug discovery

Background:

• Artificial intelligence (AI) is crucial for drug discovery but faces adoption barriers due to complex command-line tools and overfitting risks.
• Existing virtual screening (VS) software often requires coding expertise and lacks rigorous validation, leading to unreliable results.

Purpose of the Study:

• To introduce PyRMD Studio, an updated open-source VS software with a graphical user interface (GUI).
• To enhance accessibility for nonexpert users in AI-driven drug discovery workflows.
• To improve the reliability and efficiency of virtual screening.

Main Methods:

• Development of a comprehensive GUI for Linux and Windows, enabling ligand-based and structure-based VS.
• Implementation of a Butina clustering-based validation strategy to separate training and test sets, preventing data leakage.
• Code optimization and vectorization for increased computational efficiency.

Main Results:

• PyRMD Studio provides a user-friendly GUI, democratizing access to AI-powered VS for users without coding skills.
• The Butina clustering validation strategy ensures reliable performance estimation by mitigating overfitting and data leakage.
• Optimized code resulted in an approximate 3.3-fold increase in screening speed compared to the previous version.

Conclusions:

• PyRMD Studio offers a streamlined and reliable toolkit for high-throughput virtual screening campaigns.
• The software lowers the barrier to entry for AI in drug discovery, promoting wider adoption.
• Enhanced accessibility, scientific rigor, and computational efficiency make PyRMD Studio a valuable resource for researchers.