Reproduce “SENCA-st: Integrating Spatial Transcriptomics and Histopathology with Cross Attention Shared Encoder for Region Identification in Cancer Pathology” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shanaka Liyanaarachchi, Chathurya Wijethunga, Shihab Aaquil Ahamed, Akthas Absar, Ranga Rodrigo
The SENCA-st deep learning architecture integrates spatial transcriptomics and histopathology with a shared encoder and neighborhood cross-attention to identify functional regions in cancer pathology. It achieved a mean Adjusted Rand Index of 0.304 on the HER2ST breast cancer dataset, exceeding prior methods, and successfully correlated identified regions with known marker genes in Squamous Cell Carcinoma.
Key insights
- A balanced integration of spatial transcriptomics and histopathology, rather than biased approaches, is critical for accurately identifying functional regions, especially those with subtle functional differences masked by structural similarity.
- Neighborhood cross-attention effectively captures and leverages the interplay between spatial gene expression and tissue morphology to highlight regions that are structurally alike but functionally distinct.
- Combining local feature learning (via reconstruction) with global contrastive learning enables robust noise reduction in spatial transcriptomics while preserving fine-grained, biologically meaningful information.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.