Research Article

Coupling tissue morphology to gene programs across human cancers

First Author¹, Second Author¹, Third Author², Corresponding Author^1,*

¹ Department of Automation, Tsinghua University, Beijing, China · ² Collaborating Institute, City, Country

Paper Code Data Cite

Abstract

Histological morphology and molecular state are two complementary readouts of a tumour, yet how specific tissue patterns couple to gene programs remains poorly mapped. Here we learn a morphology–molecule coupling (MMC) network from paired whole-slide images and transcriptomic profiles, linking pathology foundation-model features to individual genes through comparable association weights. Spectral clustering organises the joint network into refined modules, which we annotate with Hallmark gene sets and vision–language morphology semantics. The resulting atlas recovers known axes such as Wnt–β-catenin signalling in glandular tumour epithelium and epithelial–mesenchymal transition in stromal and serrated architecture, and exposes morphology-anchored candidates for further study. This explorer provides interactive access to the network for each cancer type.

2Explore

No selection

Hover preview Click focus Double-click expand Drag pan Scroll zoom

Associations

Gene	Morphological feature	Module	Sign	Weight

Refined modules

Citation

@article{mmc2026,
  title   = {Coupling tissue morphology to gene programs across human cancers},
  author  = {Author, First and Author, Second and Author, Third and Author, Corresponding},
  journal = {Journal Name},
  year    = {2026},
  doi     = {10.xxxx/xxxxx}
}