IVS2vec: A tool of Inverse Virtual Screening based on word2vec and deep learning techniques

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Inverse virtual screening (IVS) is crucial for early-stage drug discovery, identifying biologically active molecules.
• Existing methods can be computationally intensive and may not cover all potential targets.

Purpose of the Study:

• To develop a novel, efficient prediction model for inverse virtual screening.
• To classify protein candidates based on their binding likelihood with a query molecule.
• To enhance drug discovery efficiency and safety.

Main Methods:

• Utilized Word2vec, a natural language processing technique.
• Employed a dense fully connected neural network (DFCNN) algorithm.
• Developed a binary classification model named IVS2vec.

Main Results:

• IVS2vec accurately classifies proteins into high and low binding possibility subsets.
• The model successfully identified potential therapeutic targets and targets linked to adverse drug reactions.
• IVS2vec demonstrated superior speed and performance compared to Autodock vina in reverse target searching.

Conclusions:

• IVS2vec offers a fast and effective approach for inverse virtual screening.
• The model aids in drug repurposing and improving medication safety.
• IVS2vec shows significant potential for large-scale chemical database analysis in drug development.