About HistoAtlas

While analyzing large cohorts such as TCGA, I was struck by how difficult it is to explore morphology systematically across cancers. Each whole-slide image contains millions of cells and complex spatial organization, yet we rarely represent a slide with an interpretable description of its tissue structure. I kept coming back to a simple question: why can't every slide be summarized by a histomic fingerprint capturing orthogonal aspects of the tumor microenvironment: its cellular composition, spatial organization, nuclear morphology, and tissue architecture?

Today, histomic features are computed in many studies, but they remain scattered across papers and pipelines, rarely organized into a coherent, searchable framework. As a result, researchers still lack a way to navigate the morphological landscape of cancer or to compare findings across cohorts, centers, and staining conditions. Biology has atlases for genes, proteins, and cells, but not for tissue organization.

HistoAtlas is an attempt to build that atlas: a structured way to explore cancer morphology at scale and connect tissue organization with patient outcomes and molecular signatures, enabling translational researchers to systematically discover and evaluate pathology-based biomarkers.

@misc{bannier2026histoatlaspancancermorphologyatlas,
  title         = {HistoAtlas: A Pan-Cancer Morphology Atlas Linking Histomics to Molecular Programs and Clinical Outcomes},
  author        = {Pierre-Antoine Bannier},
  year          = {2026},
  eprint        = {2603.16587},
  archivePrefix = {arXiv},
  primaryClass  = {q-bio.QM},
  url           = {https://arxiv.org/abs/2603.16587}
}