Molecule-resolved inference reveals the spatial organization of transcriptional programs

FICTURE2 infers the transcriptional program behind each individual RNA molecule using gene identity and spatial context, learning spatial patterns from millimeter-scale transitions to micron-scale subcellular localization.

• Why it matters: Spatial transcriptomics often aggregates RNA into cells or pixels, which can hide molecular-level functional dynamics and blur how gene regulation maps to phenotype.
• What they did: The method uses a statistical, segmentation-free model to assign each molecule to a generating program, supporting both unsupervised and reference-guided inference.
• The result: It learns coexisting spatial distributions that jointly reveal cellular identity and physiological activity, and it is orders of magnitude more efficient for large-scale spatial datasets across diverse tissue types.

AI summary of the abstract · may contain errors · not medical advice · read the paper.