Classification of cytochrome p(450) activities using machine learning methods.

Area of Science:

• Pharmacology
• Computational Chemistry
• Drug Discovery

Background:

• The cytochrome P450 (CYP) system is crucial for drug metabolism and xenobiotic detoxification.
• Understanding structural requirements for CYP isoform interaction is vital for early drug discovery and safety assessment.
• Major isoforms like CYP 1A2, CYP 2D6, and CYP 3A4 are key targets for drug interaction studies.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting drug interactions with CYP 1A2, CYP 2D6, and CYP 3A4.
• To identify physicochemical features associated with drug interactions with these major CYP isoforms.
• To provide tools for virtual screening and lead optimization in drug discovery.

Main Methods:

• Compilation of a dataset of 335 structurally diverse drug compounds with classified interactions (substrate, inhibitor, or any) with CYP 1A2, CYP 2D6, and CYP 3A4.
• Application of various machine learning algorithms including k-nearest neighbors, decision trees (CHAID, CRT), random forests, artificial neural networks, and support vector machines (RBF, homogeneous polynomials).
• 10-fold cross-validation to assess model performance and identify relevant physicochemical features.

Main Results:

• Machine learning models achieved high predictive accuracy, with corrected classification rates ranging from 81.7% to 92.9% across the studied CYP isoforms.
• Models demonstrated strong performance in predicting drug interactions, even with 10-fold cross-validation.
• Physicochemical features relevant to CYP interactions were discussed and compared with existing literature.

Conclusions:

• The developed machine learning models effectively predict drug interactions with major CYP isoforms.
• These models offer valuable insights into the structural determinants of CYP interactions.
• The models can serve as sensitive tools for virtual screening and optimizing toxicological profiles in drug discovery.