Tessera learns direction-sensitive spatial transcriptomics representations
Original title: Spatially Informed and Direction Sensitive Representation Learning for Spatial Transcriptomics
The method claims better representation quality than existing spatial foundation models and zero-shot use on tissues up to 16 times larger.
Tessera is a spatial representation-learning framework for spatial transcriptomics. It pairs spatially informed gene selection with a geometric bias that places each cell in its local tissue neighborhood. Relative cell-cell modeling is translation-invariant but keeps direction, so neighbors with similar distances can still differ by spatial context and cell state. The authors report stronger representation quality across held-out tissues and technologies and zero-shot generalization to contexts up to 16 times larger than those seen in training.
- Existing methods are said to drop direction or depend on absolute coordinates.
- Evaluation covers held-out tissues and multiple technologies.
- The claimed scale-up is up to 16 times the training context.
From the original article
1 min read at arxiv.orgCellular states are shaped by both intrinsic gene expression programs and the spatial microenvironment in which cells reside. Spatial transcriptomics makes this context measurable, but effective representation learning requires capturing both spatially informative molecular signals and intercellular organization.
By Azim Dehghani Amirabad · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.