Predicting neoadjuvant breast cancer therapy response using BRIDGE from tumor transcriptomics and histopathology - PubMed
an hour ago
- #neoadjuvant therapy
- #AI in oncology
- #breast cancer
- BRIDGE is a computational framework that predicts pathological complete response (pCR) to neoadjuvant breast cancer therapy by deconvolving the pre-treatment tumor transcriptome to estimate molecular subtype composition.
- It outperforms established commercial signatures like Oncotype DX, MammaPrint, and ROR-S in ER+/HER2- tumors, achieving high ROC-AUC scores across different breast cancer subtypes.
- BRIDGE-Slide extends this capability to histopathology slides by using deep learning to infer transcriptomics, offering a potential fast and low-cost biomarker for neoadjuvant response prediction.
- The framework shows generalizability to different treatment regimens, including immune checkpoint blockade in ER+/HER2- tumors, and demonstrates biological interpretability through spatial transcriptomics validation.
- Validated across multiple independent datasets, BRIDGE represents a biologically grounded tool with significant implications for precision medicine in breast cancer therapy.