Google in collaboration with the Rubin Lab at Harvard provide a fantastic overview into high-content screening as a tool for understanding disease phenotypes. Highlighting the use of machine learning as a resource for identifying patients with spinal muscular atrophy (SMA) and those without. Many of the methods comprehensively discussed in this paper are specifically important in drug repurposing such as the use of convolutional neural networks (CNNs), deep neural networks (DNNs), and individual cell-based focus scoring. Additionally, this paper delves into many of the nuances of high-content screening including plate-based imaging effects, batch-to-batch variability, and data reduction technique (i.e. t-SNE, subsampling cell populations).
https://journals.sagepub.com/doi/10.1177/2472555219857715
