The Compositional Perturbation Autoencoder (CPA) is a deep generative framework to learn effects of perturbations at the single-cell level.
Deep probabilistic analysis of single-cell and spatial omics data