High throughput microscopy generates high-dimensional data that are far from straightforward to analyze. I will describe our work in using deep neural networks to derive qualitative descriptors (e.g. subcellular localization of a fluorescent protein, or tissue of a histopathology image) and quantitative features (such as abundance of a tagged protein in cell membrane) from images. We apply these ideas to the publicly available GTEX tissue histology dataset, and yeast GFP collection micrographs.