Automated Prediction of Chemotherapy Response in PDX Models via Multi-Modal Data Fusion and Graph Neural Networks

**Abstract:** Patient-Derived Xenograft (PDX) models represent a significant advancement in preclinical drug evaluation, offering a clinically relevant platform for predicting chemotherapy response. However, manual assessment of PDX data – comprising tumor morphology, genomic profiles, and drug response curves – remains time-consuming and prone to subjectivity. This paper proposes a novel, fully automated framework leveraging multi-modal data fusion and graph neural networks (GNNs) to predict chemotherapy response in PDX models with enhanced accuracy and efficiency. The system integrates image analysis, genomic data processing, and response curve fitting into a unified model that learns complex relationships between molecular features and drug efficacy, paving the way for personalized cancer treatment selection.