Fig. 3: Detecting morphological variability during zebrafish embryogenesis and deviation from normal development.
From: Uncovering developmental time and tempo using deep learning
a, Siblings at the same nominal age may display a wide variety of morphologies due to differences in developmental tempo or differences in expression of features. b, Percentages of predicted developmental stages when comparing 77 embryos at 0 min (green), 400 min (blue) and 800 min (purple) after acquisition start with a reference embryo set with images of embryos at the age of 0.2–24.2 hpf. c, A decrease in average similarity values of a single embryo indicates phenotypic deviation, for example, due to spontaneous maldevelopment. The sum of average similarity z-scores for seven embryos is shown, each represented by a colored line. Images of a normal (top) and a defectively (bottom) developing embryo are shown at corresponding acquisition timepoints. d, Predicted stages of abnormally developing embryos (red) frequently fall outside of the normal range of predicted stages for 14 reference embryos (underlying density plot). n (maldeveloping embryos) = 7, n (normally developing embryos) = 14. On the right, a normal embryo at 17 hpf, a dissociating embryo and an embryo with dorsal–ventral patterning defects, each at 24.2 hpf, are shown with reference to the endpoints of the corresponding analyzed developmental trajectories. Scale bars, 500 μm.
