iCYP-MFE

iCYP-MFE predicts potential inhibitors of human cytochrome P450 (CYP) isoforms (1A2, 2C9, 2C19, 2D6, and 3A4) to support assessment of drug metabolism and adverse drug–enzyme interactions.

Key Features:

• Multitask learning with fingerprint embedding: Employs a multitask learning strategy that integrates molecular fingerprint-embedded features to enable simultaneous prediction of inhibitory activity across CYP isoforms.
• Isoform-specific prediction models: Comprises five prediction models tailored to each CYP isoform (1A2, 2C9, 2C19, 2D6, 3A4).
• Performance optimization: Models are designed to optimize performance metrics including area under the receiver operating characteristic curve (AUC-ROC) and area under the precision-recall curve (AUC-PR).
• Reported metrics: Achieves an AUC-ROC of 0.93 for CYP2D6-inhibition prediction and an AUC-PR of 0.92 for CYP1A2-inhibition prediction.
• Comparative evaluation: Comparative analysis reports that iCYP-MFE outperforms state-of-the-art methods in three tasks, matches one, and is less effective in one.
• Substructural analysis: Includes preliminary substructural analysis to provide insights into structural contributors to CYP-inhibitory activity.

Scientific Applications:

• Virtual screening: Prioritizes compounds for virtual screening as potential inhibitors of CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4.
• Drug discovery and development: Supports early-stage identification of compounds with CYP-inhibitory potential to inform lead selection and optimization.
• Adverse drug reaction assessment: Aids identification of compounds that may cause adverse drug reactions via CYP inhibition.
• Safety profiling: Informs the design and selection of compounds to reduce CYP-mediated safety liabilities in pharmaceutical development.

Methodology:

Uses multitask learning with molecular fingerprint-embedded features, five isoform-specific prediction models, comparative evaluation against state-of-the-art methods, optimization/assessment by AUC-ROC and AUC-PR metrics, and preliminary substructural analysis.