VirTues: A Foundation Model for Spatial Proteomics Published in Nature
DeryaTR_ · x · 2026-08-09
A new virtual tissue foundation model, VirTues, has been published in Nature. It is a general-purpose foundation model for spatial proteomics that learns multi-scale representations directly from multiplex imaging data.
From a single pretrained backbone, VirTues supports various tasks including cell segmentation, niche annotation, and zero-shot annotation across datasets. In triple-negative breast cancer, biomarkers derived from VirTues successfully predicted chemo-immunotherapy response and stratified survival, outperforming current clinical stratification schemes.
Related event: Nature Publishes VirTues: A Foundation Model for Spatial Proteomics(2 posts)→
More from Research
- LLaDA2.2 Released: Enabling Flexible Text Editing for Diffusion Language Models — jiqizhixin · 2026-08-09
- AI's Most Important Benchmarks Are the Ones No One Is Hearing About, Says Pedro Domingos — pmddomingos · 2026-08-09
- Stanford Nature Study: Intensive AI Companion Use Linked to Lower Well-Being — Diyi_Yang · 2026-08-09
- New Energy-Guided Flow Matching Boosts Image Gen Efficiency, FID to 1.55 — burny_tech · 2026-08-09
- Laplace Transform: The Mathematical Foundation Behind Mamba Beating Transformers — BLUECOW009 · 2026-08-09
- Test Shows Multi-Generation and Self-Evaluation Effectively Improve SLM Summaries — SpecialNothingness · 2026-08-09