PDRP

PDRP predicts drug responses for EGFR‑mutated lung cancer patients by combining patient-specific molecular dynamics simulations and an XGBoost classifier to generate individualized predictions for first-generation FDA-approved EGFR tyrosine kinase inhibitors such as Gefitinib and Erlotinib.

Key Features:

• Molecular Dynamics Simulations: Patient-specific MD simulations of mutated EGFR generate molecular-level geometric features of the drug–target binding site.
• Machine Learning Integration: An XGBoost classifier integrates molecular features with demographic and clinical information (DCI), with DCI reported to have minimal impact compared to molecular data.
• Feature Set: Inputs include demographic and clinical information, geometrical properties of the drug–target binding site, and binding free energy of the drug–target complex derived from MD simulations.
• Targeted Drugs and Disease: Predicts responses to first-generation FDA-approved EGFR tyrosine kinase inhibitors Gefitinib and Erlotinib in lung cancer patients harboring EGFR mutations.
• Performance Metrics: Reported performance on four-class response prediction: 97.5% accuracy, 93% recall, 96.5% precision, and 94% F1-score.

Scientific Applications:

• Drug response prediction in EGFR‑mutated lung cancer: Forecasts individual responses to Gefitinib and Erlotinib to inform personalized treatment planning.
• Mechanistic interpretation of resistance: Uses MD-derived geometric features and binding free energies to analyze how EGFR mutations affect drug binding and potential resistance mechanisms.
• Methodological adaptability: Could be adapted to other cancer targets and therapies using the same MD plus machine learning paradigm.

Methodology:

Patient-specific MD simulations of mutated EGFR were conducted to extract geometric properties of the binding pocket and binding free energies, which were combined with demographic and clinical data and input to an XGBoost classifier for four-class drug-response prediction.