HORIZON model learns tissue dynamics while ignoring rigid motion
A new arXiv paper introduces horizontal flow matching, a method that models how tissue organization changes over time without mistaking rotation or translation of a sample for real cell movement.
In plain words
Spatial transcriptomics captures snapshots of tissue at different times, but each snapshot sits in its own coordinate frame, so a rotated or shifted section can look like cells have moved. The authors' method, called horizontal flow matching, learns changes in tissue structure, molecular state, and cell abundance while treating shared rigid motion as cost-free. Their model, HORIZON, can then generate a tissue's evolution from one starting snapshot.
What was announced
The method, described in a cross-posted arXiv abstract, frames the problem as a continuity–reaction equation whose action assigns no cost to shared rigid motion. A horizontal projector removes this component from prediction errors during training and from predicted velocities during generation. The authors also show an equivalent endpoint problem that jointly estimates correspondence and relative pose in Wasserstein–Fisher–Rao geometry, producing analytic paths whose velocity and growth serve as regression targets; regressing these targets recovers the horizontal dynamics under ideal population conditions.
HORIZON itself uses an equivariant transformer to predict velocity and growth from the current tissue, allowing continuous generation from a single initial snapshot. In the reported experiments, alignment improved reconstruction on a simulated tissue observed in perturbed coordinate frames, and projection reduced rigid drift. Across four spatial transcriptomics time courses, the authors report that HORIZON reconstructed tissue geometry, gene expression, and cell abundance more accurately than existing spatial and unbalanced trajectory methods, and interpolated held-out developmental stages. These results come from the paper's own abstract and have not been independently verified.
What it does not show
- Evidence is limited to the paper's abstract; full experimental details, datasets, and baselines are not available in the supplied passages.
- Reported performance claims are the authors' own and have not been independently confirmed.
Sources
Source-support review completed 20 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.
Coverage: arXiv: abstract.
Updates and corrections
No corrections so far.
First published 20 Nov 2026
New to the subject? AI from the Ground Up starts from the beginning.