Reproduce “Screen Them All: High-Throughput Pan-Cancer Genetic and Phenotypic Biomarker Screening from H&E Whole Slide Images” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yi Kan Wang, Ludmila Tydlitatova, Jeremy D. Kunz, Gerard Oakley, Bonnie Kar Bo Chow, Ran A. Godrich, Matthew C. H. Lee, Hamed Aghdam and 10 more
Paige researchers introduced OmniScreen, a high-throughput, pan-cancer AI system that predicts 1,637 genetic and phenotypic biomarkers directly from routine H&E whole slide images. The system achieved a mean AUC of 0.90 for 93 top-performing genomic alterations and demonstrated robust accuracy for clinically actionable targets and genomic instability markers, showing potential for significant cost savings and improved clinical workflow efficiency.
Key insights
- AI can identify subtle pan-cancer morphological patterns in H&E WSIs that correlate with a broad spectrum of genomic alterations, pathways, and genomic instability, often beyond human visual perception.
- A unified, pan-cancer model approach is more effective for predicting low-prevalence biomarkers compared to fragmented, single-target models, offering greater efficiency and scalability.
- OmniScreen can detect 'pathway-active' phenotypes where multiple oncogenes converge on shared signaling cascades, providing biological insights into the downstream morphological effects of molecular dysregulation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.