Fig. 2: Automated analysis of fish developmental temperature dependence using Twin Networks.
From: Uncovering developmental time and tempo using deep learning
a–f, Analysis of zebrafish and medaka embryo development at various temperatures. a,d, Schematic for age estimation of zebrafish (a) and medaka (d): image of an embryo at a given timepoint (y hpf), raised at the temperature of interest (x °C), is compared with all timepoints (three examples are shown) at the reference temperature. Developmental age is assigned by the highest cosine similarity (φ). Scale bars, 500 μm. b,e, Developmental age estimation for zebrafish (b) embryos at 26.5 °C, 28.5 °C and 31.5 °C (n = 209, 126 and 130, respectively) and medaka (e) embryos at 26.0 °C, 28.0 °C and 31.0 °C (n = 47, 46 and 21), respectively. Error envelopes represent two times the median absolute deviation (MAD) over the embryos and are shown together with the corresponding linear fit (solid line). c,f, Natural logarithm of the estimated growth rates for zebrafish (c) and medaka (f) at various temperatures. Error bars represent 99.99% confidence intervals from bootstrapping with 100 repetitions around the estimated slope of the linear fit to the data shown in Extended Data Figs. 3 and 4. Blue shading shows the Arrhenius range; the apparent activation energy is stated.
