A deep generative model of 3D single-cell organization

We introduce a framework for end-to-end integrative modeling of 3D single-cell multi-channel fluorescent image data of diverse subcellular structures. We employ stacked conditional beta-variational autoencoders to first learn a latent representation of cell morphology, and then learn a latent representation of subcellular structure localization which is conditioned on the learned cell morphology. Our model is flexible and can be trained on images of arbitrary subcellular structures and at varying degrees of sparsity and reconstruction fidelity. We train our full model on 3D cell image data and explore design trade-offs in the 2D setting. Once trained, our model can be used to predict plausible locations of structures in cells where these structures were not imaged. The trained model can also be used to quantify the variation in the location of subcellular structures by generating plausible instantiations of each structure in arbitrary cell geometries. We apply our trained model to a small drug perturbation screen to demonstrate its applicability to new data. We show how the latent representations of drugged cells differ from unperturbed cells as expected by on-target effects of the drugs.

Automated summary

We propose a framework for integrative modeling of 3D single-cell multi-channel fluorescent image data, allowing for the representation and localization of diverse subcellular structures. Our model utilizes stacked conditional beta-variational autoencoders to learn latent representations of cell morphology and subcellular structure localization. It is capable of training on different subcellular structures with varying sparsity and reconstruction fidelity. The trained model can predict plausible locations of structures in cells that were not imaged, as well as quantify the variation in the location of subcellular structures. We demonstrate the applicability of our model to new data in a small drug perturbation screen, showing expected differences in the latent representations of drugged and unperturbed cells.