PlayMolecule BindScope: large scale CNN-based virtual screening on the web.

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Virtual screening is crucial for structure-based drug discovery, reducing time and costs.
• Deep learning methods show superior performance over classical scoring functions in discriminating protein-ligand interactions.

Purpose of the Study:

• To introduce BindScope, a web application for large-scale compound classification.
• To leverage deep convolutional neural networks for active-inactive compound identification.

Main Methods:

• Development of BindScope, a web application utilizing deep convolutional neural networks.
• Implementation of large-scale active-inactive classification of chemical compounds.
• Integration into the PlayMolecule.org web application suite.

Main Results:

• BindScope achieves performance comparable to state-of-the-art virtual screening pipelines.
• Enables screening of hundreds of compounds simultaneously.
• Provides interactive visualization of screening results.

Conclusions:

• BindScope offers an efficient and effective tool for structure-based drug discovery.
• The application facilitates rapid identification and analysis of potential drug candidates.
• Deep learning-based virtual screening shows significant promise in accelerating drug development.