Kraus, O.Z. et al. Mol. Syst. Biol. 13, 924 (2017).

Automated image analysis methods can be useful for improving the quality and speed of quantitative assessment of images. Deep-learning methods have emerged as powerful tools for improving image analysis beyond what is capable with existing computational pipelines. Kraus et al. build on this trend by developing DeepLoc, a deep convolutional neural network to analyze images of yeast cells. A benefit of this approach is that extensive training is not needed to use the model on new imaging data sets, and so the model is readily transferable. The researchers demonstrate that DeepLoc offers improved determination of protein subcellular localization in yeast relative to existing tools. They also validated that the tool works on data sets that are from different screens than those used for training.