Drug sensitivity prediction across cancer types using graph isomorphism networks and biological pathway features: A dual-branch deep learning approach

Molecular Fragment-Based Graph Isomorphism Networks for Interpretable Prediction of Synergistic Drug Combinations

Results indicate that FragSyn, through the synergistic design of fragment-level representation and cellular context awareness, provides an effective and interpretable new approach to synergistic drug combination prediction.

[Prediction and verification of therapeutic drugs for triple-negative breast cancer using a knowledge graph-based drug repurposing model].

Essentiality-driven prediction of anticancer drug responses in preclinical and clinical contexts

DrGee is presented, an essentiality-centered platform that infers drug sensitivity solely from gene expression profiles, and the built-in DeepEEAA model integrates gene expression, gene essentiality, drug-protein affinity, and drug-gene associations to quantitatively predict IC50 values.

Graph neural networks for multi-scale gene-drug interaction modeling in cancer: applications, challenges, and therapeutic opportunities for breast cancer drug resistance

Breast cancer drug resistance remains a major clinical challenge driven by complex genetic, signaling, and microenvironmental interactions. Conventional machine learning and deep learning represent genes, drugs, and patients as independent feature vectors, limiting their ability to capture biological relationships governing therapeutic response. Graph neural networks have emerged as a powerful paradigm by modelling biological systems as interconnected networks rather than isolated entities. This review synthesizes recent advances in graph neural network based multi-scale modelling of gene-drug interactions across cancer research, emphasizing translational relevance to breast cancer drug resistance. Although many architectures originate from pan-cancer or methodological studies, their potential for breast cancer is critically assessed while distinguishing models directly validated from those requiring adaptation. Recent architectures, including hierarchical, heterogeneous, knowledge graph-guided, explainable, and graph-augmented models, demonstrate strong predictive performance across oncology tasks. These studies computationally nominated potential targets such as FAK, FLT3, COX8A, SEC61G, and CYP27B1, though predictions require experimental validation in breast cancer resistance models. The field has leveraged established synthetic lethal relationships, such as BRCA1/PARP, as benchmarks for GNN-based discovery frameworks. Despite encouraging progress, current evidence remains largely retrospective, benchmark-based, or preclinical. Cross-cohort heterogeneity, limited interpretability, and scarce breast cancer-specific resistance validation represent central limitations. Future integration with spatial transcriptomics, multimodal omics, and federated learning may improve precision oncology, but rigorous biological validation and interdisciplinary collaboration are essential for clinical implementation.

Predicting circRNA-Drug Sensitivity Using Integrated Hybrid Graph and Molecular Features.

Case studies further validated predicted circRNA-drug associations, including cisplatin, enzalutamide, and sorafenib, against published experimental findings, demonstrating HMCDSP's capacity to reveal clinically relevant biomarkers and inform personalized therapeutic strategies.

Prediction of Drug Target Interactions Based on Artificial Intelligence

This review systematically summarizes the latest developments in heterogeneous graph neural networks, protein language models, and generative artificial intelligence, pointing out the problems currently being addressed in research such as data sparsity and cold start as well as the manifestations of general machine learning challenges.