Extended Data Fig. 3: Average droplet size and geometry is robust to thresholding of images.

a, To test the robustness of the image segmentation to thresholding, we first calculated the threshold as described in the main text methods (two standard deviations above mean nuclear intensity at time 0). We then multiplied the threshold by a factor ranging from 0 to 5 (‘Factor change’) and calculated what percentage of each image was above the threshold and would be identified as a droplet (‘Fraction of image’). Each colored curve corresponds to a single image containing one nucleus (N=18 images total) after 105 minutes of activation; black dashed line reports the average over all images. Trivially, the entire image is above the threshold at a factor value of 0 and decreases quickly. However, as the factor approaches 1 and higher values (that is, the chosen threshold) the fraction of the image segmented decreases only slowly (see inset), suggesting that segmentation is robust with respect to the threshold above these values. b, The identical procedure was applied to the same images to evaluate the dependence of the average droplet radius on threshold, as demonstrated in inset example images. Similar to the fraction of image metric, the average droplet radius was robust above a factor of approximately 0.75, suggesting that our results are not sensitive to the precise threshold.