Figure 1.

Multicolor skin cell images captured from live zebrafish larvae as input for machine learning.

(A) The palmskin transgenic constructs. (B) Schematic cross-section of a palmskin larva. Upon a 2-h pulse treatment of tamoxifen (Tam) and doxycycline (Dox), the superficial epithelial cells covering the trunk surface area in the bi-layered epidermis are barcoded in multicolor. SEC, superficial epithelial cell. BEC, basal epithelial cell. (C) Schematic timeline illustrating the Dox and Tam treatment protocol and the imaging schedule. dpf, days post-fertilization. (D) Schematic illustration of the local palmskin images used for machine learning. (E) Representative palmskin images captured from fish of varying growth sizes. Scale bar, 50 μm. (F) K-means clustering of the trunk surface area of 361 fish. Each dot represents an individual fish, with small-sized (S) fish shown in red, medium-sized (M) fish shown in green, and large-sized (L) fish shown in blue. Dotted lines indicate the cut-offs for different size categories. (G) Schematic flowchart of the inputs, machine learning models, and growth size classification outputs.