Sophont's Holon-S1 predicts genes from pathology slides, beats DeepSpot-M and Path2Space
iScienceLuvr · x · 2026-10-21
Sophont launched Holon-S1, a platform that predicts genes and biomarkers directly from routine H&E pathology slides.
- Beats two 2026 models on spatial gene-activity prediction (median per-gene correlation): 0.49 vs 0.21 against DeepSpot-M in breast cancer, 0.31 vs 0.11 in unseen cervical cancer; 0.34 vs 0.17 against Path2Space, which is breast-only while Holon-S1 is pan-cancer
- Has surfaced potentially novel biomarkers in triple-negative breast and colon cancer
- Evaluated across 16 organ types, 15+ scanners, and both FFPE and fresh-frozen tissue
- Unlike black-box pathology AI, it maps visible genes and biological programs across tumors
Research use only for now; the team says prospective validation could make personalized cancer care more accessible.
Related event: Holon-S1 Predicts Spatial Gene Expression Directly from Pathology Slides(4 posts)→
More from Apps
- LukeW maps the five UI problems every AI-native company faces with agents — LukeW · 2026-10-21
- Pinokio ditches the algorithm for a pure follow-based feed for AI apps — cocktailpeanut · 2026-10-21
- Hamlet taps AssemblyAI to turn 100K+ local government meetings into civic intelligence — AssemblyAI · 2026-10-21
- Reporter's one Gmail feature ask: make Gemini search as good as searching Gmail via ChatGPT — firstadopter · 2026-10-21
- Google Labs launches Dreambeans, an AI app that brews your connected apps into daily stories — GoogleLabs · 2026-10-21
- Radical Numerics' Omnii-1 designed world-first AI phages given to a human patient — JosephJacks_ · 2026-10-21