What is it about?
Spatial transcriptomics measures gene activity across tissues while preserving the location of each measurement. Many existing methods group these tissue locations into clusters, but they often overlook how genes are related to each other inside each location. In this work, we show that considering these gene–gene relationships yields more accurate tissue clustering. We introduce a new method, STING, that models both the spatial arrangement of the locations and the relationships between genes within each location. Across various datasets, this approach enhances clustering performance and also reveals gene interactions that may be biologically significant.
Featured Image
Photo by Maria Kovalets on Unsplash
Why is it important?
This work incorporates gene–gene relationships into spatial transcriptomics clustering for the first time. As datasets grow larger and richer, these internal gene relations may become increasingly important. By modelling them directly, we can improve cell clustering and reveal gene relationships that may offer new biological insights.
Perspectives
I hope that the methodology we used to integrate two different networks can be useful to spatial transcriptomics and other research fields. I also hope that any future research that improves this method would also be beneficial to multiple research fields.
Atishay Jain
Brown University
Read the Original
This page is a summary of: Improved Spatial Transcriptomics Clustering with Nested Graph Neural Networks, October 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3765612.3767241.
You can read the full text:
Contributors
The following have contributed to this page
