Hi Product Hunt! I'm Pierre-Antoine, the maker of HistoAtlas. The problem: Cancer pathology slides contain rich morphological information, but there's no easy way to systematically explore how tissue architecture connects to molecular data and patient outcomes across cancer types. Existing tools like cBioPortal focus on genomics. What I built: HistoAtlas takes 6,745 diagnostic slides from TCGA, extracts 38 quantitative features (immune cell densities, tumor morphology, spatial organization), and precomputes every statistical association with genes, mutations, pathways, and survival. The result is a fully interactive web atlas you can explore instantly. Tech stack: Astro + React 19 (static-first), Tailwind v4, Deck.gl for the UMAP, Recharts for charts. The entire API is pre-generated static JSON: no backend, no database, sub-100ms loads. The analysis pipeline is Python (Cox PH, Spearman, K-means, Benjamini-Hochberg FDR correction). This is a solo project I built because I wanted this tool to exist for my own research. Happy to answer any questions about the data, methods, or tech! Try it: https://histoatlas.com

HistoAtlas

Pan-cancer morphology atlas from 6,745 TCGA slides

Pan-cancer morphology atlas from 6,745 TCGA slides