Artificial intelligence in virtual screening: Models versus experiments.

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Artificial intelligence in drug discovery

Background:

• Traditional drug discovery relies on experimental high-throughput screening (HTS), which is resource-intensive for large chemical libraries.
• Computational screening methods offer a way to narrow down the search space for potential drug candidates.
• Machine learning (ML) and deep learning (DL) are increasingly utilized in drug discovery pipelines.

Purpose of the Study:

• To review machine learning and deep learning-based scoring functions for drug discovery.
• To highlight successful applications of ML and DL in identifying lead compounds.
• To provide an overview of computational approaches for classification and ranking in drug discovery.

Main Methods:

• Review of existing literature on ML and DL scoring functions in drug discovery.
• Analysis of studies employing ML and DL for compound classification and ranking.
• Focus on models with experimentally validated lead compounds.

Main Results:

• ML and DL models demonstrate effectiveness in classifying and ranking compounds.
• Successful identification of lead compounds using ML and DL approaches.
• Experimental validation confirms the utility of ML/DL-identified compounds.

Conclusions:

• ML and DL-based scoring functions are powerful tools for efficient drug discovery.
• Computational screening significantly reduces the time and cost associated with identifying drug candidates.
• The integration of ML and DL accelerates the identification of validated lead compounds for further development.