Improved genome-scale multi-target virtual screening via a novel collaborative filtering approach to cold-start...

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Traditional drug discovery often uses a one-drug-one-gene approach with limited success.
• Polypharmacology, targeting multiple proteins in disease networks, is a promising new paradigm.
• Existing virtual screening methods are largely designed for single targets, not polypharmacology.

Purpose of the Study:

• To develop a novel theoretical framework and algorithm for genome-scale multi-target virtual screening.
• To address the challenge of sparse protein-chemical interaction data in drug discovery.
• To identify potential protein targets for new chemical compounds with limited interaction data.

Main Methods:

• Developed a one-class collaborative filtering technique for multi-target screening.
• Implemented interaction matrix weighting and dual regularization to handle data sparsity.
• Evaluated the algorithm using gene-specific and cross-gene family benchmarks.

Main Results:

• The proposed method effectively overcomes the sparseness of protein-chemical interaction data.
• The algorithm demonstrates superior performance compared to state-of-the-art methods in predicting interactions with multiple proteins.
• The framework is adaptable for genome-wide drug off-target prediction.

Conclusions:

• The novel algorithm provides a powerful tool for advancing multi-target drug design.
• This approach enhances the efficiency and scope of virtual screening in drug discovery.
• The method facilitates the identification of new chemical-protein interactions for therapeutic development.