Smart distributed data factory volunteer computing platform for active learning-driven molecular data acquisition.

Area of Science:

• Computational chemistry
• Artificial intelligence in drug discovery
• Data science for molecular modeling

Background:

• Drug discovery faces challenges with vast chemical spaces and limited high-quality data.
• Accelerating quantum chemistry calculations is crucial for molecular property prediction.
• Existing methods often require extensive computational resources.

Purpose of the Study:

• To present the Smart Distributed Data Factory (SDDF), an AI-driven platform for generating comprehensive molecular datasets.
• To leverage volunteer computing and machine learning to accelerate drug discovery.
• To create a scalable and cost-effective solution for developing accurate molecular models.

Main Methods:

• Utilizing volunteer computing for distributed density functional theory (DFT) calculations.
• Employing an ensemble of machine learning (ML) models for property prediction and active learning.
• Generating new molecular conformations via molecular dynamics using ML-derived forces.

Main Results:

• Developed a volunteer computing platform for accelerated DFT calculations.
• Constructed a large, public dataset of diverse ENAMINE molecules with calculated energies.
• Created an ensemble of ML models demonstrating accurate energy prediction capabilities.
• Validated the SDDF approach in reducing the need for extensive computational calculations.

Conclusions:

• SDDF combines active learning, distributed computing, and quantum chemistry for efficient molecular modeling.
• The platform offers a scalable and cost-effective solution for accelerating drug discovery.
• The generated dataset supports training and benchmarking of energy prediction models.