Prediction of chemical-protein interactions network with weighted network-based inference method

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Chemical-protein interactions (CPIs) are crucial for target identification and drug discovery.
• Experimental determination of large-scale CPIs is challenging and costly.
• In silico prediction methods offer a cost-effective and accurate alternative.

Purpose of the Study:

• To develop and evaluate novel node- and edge-weighted network-based inference (NBI) methods for CPI prediction.
• To assess the reliability and potential applications of these NBI methods in drug discovery.

Main Methods:

• Utilized two comprehensive CPI bipartite networks from the ChEMBL database.
• Applied node- and edge-weighted NBI methods for predicting CPIs.
• Validated prediction reliability using external validation sets and receiver operating characteristic curves.

Main Results:

• Achieved area under the curve values ranging from 0.73 to 0.83 in external validation, confirming prediction reliability.
• The edge-weighted NBI method successfully identified the weak-interaction hypothesis in CPI networks.
• Successfully predicted candidate targets for five approved drugs, providing molecular hypotheses and experimental evidence.

Conclusions:

• The developed NBI methods demonstrate significant potential for understanding the molecular basis of drug polypharmacology.
• These computational tools can aid in drug repositioning by identifying novel drug-target interactions.
• The study highlights the value of in silico approaches for advancing drug discovery and development.