High-content image-based drug screens can be used to identify novel compounds affecting cellular phenotypes. However, larger drug screens assessing whole phenomes with rich biological information are currently hampered by the large number of images that have to be visually assessed by a limited number of experts combined with the ambiguity of defects and variability between assessors. To validate the general applicability of EmbryoNet in linking chemical manipulations to signaling-based phenotypic defects, we performed a large-scale zebrafish screen using FDA-approved and bioactive compunds.