AI-Driven Spatial Transcriptomics Unlocks Large-Scale Breast Cancer Biomarker Discovery from Histopathology

Join Dr. Eytan Ruppin, NCI investigator in the Center for Cancer Research, as he discusses Path2Space, a new and unpublished deep learning approach that predicts spatial gene expression directly from histopathology slides.

Spatial transcriptomics (ST) is transforming our understanding of tumor heterogeneity by providing high-resolution, location-specific mapping of gene expression within tumors and their microenvironment. However, high costs have restricted the size of cohorts, limiting large-scale biomarker discovery.

With Path2Space, you can:

• predict the spatial expression of over 4,300 breast cancer genes in independent validations, thereby outperforming existing ST predictors.
• accurately infer cell-type abundances in the tumor microenvironment (TME).
• apply to over 1,000 breast tumor histopathology slides from the TCGA, characterizing their TME on an unprecedented scale, and identify new spatially grounded breast cancer subgroups with distinct survival rates.
• infer TME landscapes, enabling more accurate predictions of patients’ response to chemotherapy and trastuzumab.
• operate a transformative, fast, and cost-effective approach to robustly delineate the TME.

The CBIIT Data Science Seminar Series is dedicating its 2025 events to spotlighting the use of AI in cancer research and care. Brought to you by CBIIT and NCI’s Division of Cancer Treatment and Diagnosis AI working group, the upcoming webinars will explore a variety of questions, such as the following:

• How can we use AI for diagnosis, treatment, drug development, and omics research?
• What are the related laws and ethical considerations for AI?
• How can we empower an AI-ready cancer research community through workforce development, collaborations, and funding?

To view upcoming speakers or recordings of past presentations, visit the CBIIT Data Science Seminar Series webpage.