iCYP-MFE: Identifying Human Cytochrome P450 Inhibitors Using Multitask Learning and Molecular Fingerprint-Embedded...

Area of Science:

• Biochemistry and Pharmacology
• Computational Chemistry
• Drug Discovery

Background:

• Cytochrome P450 (CYP) enzymes are vital for metabolizing endogenous and exogenous compounds, including drugs.
• CYP inhibition is linked to adverse drug reactions and metabolic failures.
• Identifying potential CYP inhibitors is essential in drug discovery.

Purpose of the Study:

• To introduce iCYP-MFE, a computational framework for virtual screening of CYP inhibitors.
• To develop robust prediction models for human CYP isoforms 1A2, 2C9, 2C19, 2D6, and 3A4.

Main Methods:

• Utilized multitask learning with molecular fingerprint-embedded features to create five prediction models.
• Employed area under the receiver operating characteristic curve (AUC-ROC) and area under the precision-recall curve (AUC-PR) for model evaluation.
• Performed substructural analysis to understand CYP-inhibitory activity.

Main Results:

• iCYP-MFE demonstrated strong performance, outperforming state-of-the-art methods in key tasks.
• Achieved high AUC-ROC of 0.93 for CYP2D6 inhibition and AUC-PR of 0.92 for CYP1A2 inhibition.
• The framework leverages related tasks to enhance global prediction performance.

Conclusions:

• iCYP-MFE is an effective computational tool for identifying CYP inhibitors.
• The framework supports drug discovery by predicting potential adverse drug reactions early.
• An accessible web server was deployed to aid the scientific community.