SpinePy enables automated 3D spatiotemporal quantification of multicellular in vitro systems

Abstract

Organoids and stem-cell-based embryo models such as gastruloids are powerful systems to quantitatively study morphogenesis and patterning. This requires 3D analysis in reference frames that emerge dynamically over development, but are less stereotypical than in vivo. Meaningful statistical comparison and interpretation of biological and physical quantities in space and time -- such as signaling activity, gene expression or cell flows -- depend on proper quantification in these internal and dynamic coordinate systems, especially in in vitro systems that naturally exhibit larger variation. Here, we present a computational framework, packaged as a modular Python toolkit termed SpinePy, that identifies the emergent primary body axis ("spine") of individual gastruloids and constructs a local, dynamic coordinate system aligned to their morphology. SpinePy enables 3D quantification relative to evolving axes and statistical comparisons of morphodynamics, patterning, and densities across structures with varying geometries. We validate and benchmark SpinePy using both synthetic and experimental gastruloid data, providing practical insights into method performance. Using this framework, we generate 3D patterning maps from gastruloids formed with different initial cell numbers (N0). This reveals distinct patterning classes that are better explained by gastruloid volumes than by N0. While demonstrated in gastruloids, SpinePy is broadly applicable to any multicellular system where analysis relative to evolving internal axes is needed, advancing quantitative and comparative spatial biology.