A review of deep learning methods for ligand based drug virtual screening

Area of Science:

• Computational chemistry
• Pharmacology
• Artificial intelligence in drug discovery

Background:

• Drug discovery is expensive and time-consuming, with computational methods increasingly vital.
• The COVID-19 pandemic highlighted the need for faster drug and vaccine development.
• Deep learning (DL) shows significant promise in accelerating drug virtual screening.

Purpose of the Study:

• To provide a comprehensive overview of deep learning methods in drug virtual screening.
• To compare and analyze the performance of various DL models for virtual screening tasks.
• To identify challenges and future directions in computational drug discovery.

Main Methods:

• Introduction to fundamental concepts of drug virtual screening, datasets, and data representation.
• Comparative analysis of numerous common deep learning methods applied to virtual screening.
• Independent evaluation of DL model performance on large-scale ligand virtual screening datasets.

Main Results:

• Detailed comparison of deep learning model performance across different dataset sizes.
• Identification of strengths and weaknesses of various deep learning approaches for virtual screening.
• Empirical data supporting the effectiveness of specific DL models for large-scale screening.

Conclusions:

• Deep learning significantly enhances the efficiency and accuracy of drug virtual screening.
• Further research is needed to optimize DL models for complex drug discovery challenges.
• The review provides a roadmap for leveraging AI to accelerate the development of new therapeutics.