Reviewing reproducibility and replicability in 'Deconstructing gastrulation at single-cell resolution' by Stern et al. (2022).

Quantitative analysis of embryonic morphogenesis often requires handling terabyte-scale 4D microscopy datasets, posing major computational and methodological challenges. Such analyses involve complex workflows for image restoration, segmentation, and tracking of thousands of cells in 3D, sometimes assembled from multiple generic tools and custom scripts. In their study 'Deconstructing gastrulation at single-cell resolution,' the authors present an integrated computational framework for cell segmentation, tracking and modelling of cell-cell interactions during Drosophila gastrulation. Their approach combines both generic tools such as ilastik, FIJI, Cellpose, MATLAB code and tools such as ImSAnE to reconstruct whole-embryo, single-cell morphogenetic atlas with unprecedented detail. While the methodology is transparent and well-documented, its implementation relies on multiple software environments and some non-open-source components, limiting full reproducibility and generalisability. Nevertheless, their 'computational framework' exemplifies rigorous data sharing, clear documentation, and methodological transparency aligned with several community guidelines for reproducible image analysis. We discuss this work in the context of reproducibility, replicability, generalisability, transparency, validity, and robustness, emphasising how emerging open-source and popular solutions could improve accessibility and replicability of similar large-scale bioimage analysis workflows in developmental biology.