Optimal decision-making in high-throughput virtual screening pipelines

Area of Science:

• Computational chemistry
• Materials science
• Drug discovery

Background:

• Efficient screening of molecular candidates is crucial for drug discovery and materials design.
• Large search spaces and high-fidelity model costs hinder practical screening.

Purpose of the Study:

• To develop a general framework for constructing and optimizing high-throughput virtual screening (HTVS) pipelines.
• To optimally allocate computational resources among multi-fidelity models for improved efficiency.

Main Methods:

• Proposed a framework for HTVS pipelines utilizing multi-fidelity models.
• Developed an optimal resource allocation strategy based on model cost and accuracy.
• Validated the framework using simulated and real-world data.

Main Results:

• Demonstrated significant acceleration of virtual screening processes.
• Achieved acceleration without compromising predictive accuracy.
• Enabled adaptive strategies to trade accuracy for computational efficiency.

Conclusions:

• The proposed optimal HTVS framework enhances computational screening efficiency.
• The framework offers flexibility in balancing accuracy and speed for virtual screening tasks.
• This approach is valuable for accelerating discovery in fields like drug development and materials design.