Research report: A kinematic interpretation of spatial expression gradients from single-snapshot spatial transcriptomics

A computational framework can interpret spatial gene-expression gradients measured in a single tissue section as indicators of net-production direction, but only under explicit assumptions about a stationary tissue state and positive movement along an anatomical axis. Tests using a human liver atlas and mouse intestinal tissue found reproducible gradient directions, while diagnostics showed that relationships among gene gradients cannot be reliably read as a regulatory network from one sampled trajectory. The results support a splicing-free estimator of spatial expression gradients, while also defining the limits of what cross-sectional data can reveal.

The method addresses a central interpretive problem in spatial transcriptomics: a gene’s expression may vary systematically across tissue, but a spatial pattern alone does not establish what produced that pattern. The framework assumes that the position-conditioned expression field is stationary and that cells have positive mean displacement along an anatomical axis. Under these conditions, the sign of a gene’s spatial gradient equals the sign of its net production along that axis.

A positive gradient is therefore interpreted as net production in the positive direction, while a negative gradient indicates net production in the opposite direction. The statistic measured from the tissue remains a spatial gradient; the production interpretation is conditional rather than a direct temporal observation. Without an independent velocity estimate, the approach does not recover absolute production rates.

The researchers evaluated the framework in a human liver atlas containing 16 standard-Visium sections, three matched Visium HD sections and companion single-nucleus RNA sequencing. The primary cross-donor analysis used eight sections from living donors. On a strict panel of 66 genes, the mean cross-donor correlation of gradient direction was 0.87.

Across a broad panel of 15,192 genes, the corresponding mean correlation was 0.67. On prespecified top-quartile subsets, cross-platform sign agreement was 14 out of 14 genes for the M2 measurement and 13 out of 14 for M6, with 14 genes assessed per donor. Agreement across analysis pipelines was 76%.

These observations show that the measured direction of expression gradients was reproducible in the human liver evaluation across the tested donors, platforms and pipelines. Four diagnostics and a geometric argument indicated that this coupling matrix is not stably identifiable or interpretable from a single sampled trajectory. The result does not mean that genes cannot be associated, but it shows that a one-snapshot matrix cannot, on the evidence provided, establish direct regulatory relationships.

The distinction separates the observable—a reproducible spatial gradient—from its mechanistic interpretation and demonstrates why additional information would be needed for stronger claims about gene regulation. In a separate analysis of graded steatosis across four donors, GLUL, LYZ and ORM1 showed consistent spatial-gradient responses to lipid. However, cluster-robust calibration of significance was anticonservative, so the authors report these results as directional observations rather than statistically significant discoveries.

On the mouse crypt-to-villus axis, all 15 canonical markers recovered their expected directions when the coordinate was derived from anatomy. Together, the liver and mouse analyses support the framework’s applicability to imaging-based spatial transcriptomics while emphasizing the boundary of the evidence: cross-sectional data establish gradient reproducibility, not time-resolved dynamics, and the net-production interpretation applies only when the stated steady-state and positive-flow assumptions hold.

Subject of Research: Kinematic interpretation of spatial gene-expression gradients in single-snapshot spatial transcriptomics

Article Title: A kinematic interpretation of spatial expression gradients from single-snapshot spatial transcriptomics

Article References: Iešmantas, T. (2026). A kinematic interpretation of spatial expression gradients from single-snapshot spatial transcriptomics. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06707-6

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06707-6

Keywords: spatial transcriptomics, gene-expression gradients, net production, human liver atlas, Visium, single-nucleus RNA sequencing, mouse intestinal tissue, steatosis, computational genomics, tissue dynamics, kinematic, interpretation