Unsupervised discovery of dynamic cell phenotypic states from transmitted light movies

Abstract

SummaryIdentification of cell phenotypic states within heterogeneous populations, along with elucidation of their switching dynamics, is a central challenge in modern biology. Conventional single-cell analysis methods typically provide only indirect, static phenotypic readouts. Transmitted light images, on the other hand, provide direct morphological readouts and can be acquired over time to provide a rich data source for dynamic cell phenotypic state identification. Here, we describe an end-to-end deep learning platform, UPSIDE (for Unsupervised Phenotypic State IDEntification), for discovering cell states and their dynamics from transmitted light movies.UPSIDE uses the variational auto-encoder architecture to learn latent cell representations, that are then clustered for state identification, decoded for feature interpretation, and linked across movie frames for transition rate inference. Using UPSIDE, we identified distinct blood cell types in a heterogeneous dataset. From acute myeloid leukemia cell movies, we then identified stem-cell associated morphological states and their inter-conversion rates. UPSIDE opens up use of transmitted light movies for systematic exploration of cell state heterogeneity and dynamics in biology and medicine.

DeepIFC: virtual fluorescent labeling of blood cells in imaging flow cytometry data with deep learning

“…Taken together, methods such as DeepIFC able to perform virtual labeling and cell type identification solely from morphology hold promise to transform diagnosis of hematological diseases and blood processing pipelines by not having to introduce fluorescent labels during workflow, reducing costs and processing time required. Possible avenues to develop the method further include utilizing larger training datasets covering more cell types, data augmentation to improve performance on rare cell types (Luo, Nguyen et al . 2021), and a user-friendly graphical tool to use the software.…”

“…Label-free deep learning methods have been created to reconstruct fluorescent images from brightfield images (Christiansen et al 2018, Ounkomol et al 2018, Nguyen et al 2021), but to our knowledge, the reconstruction of single-cell multichannel fluorescent images in IFC data has not been proposed. Label-free cytometry methods based on segmentation and unsupervised modeling (Nguyen et al 2021), weak supervision (Otesteanu et al 2021), cytometry by time of flight (CyTOF) (Hu et al 2020) and time-stretch microscopy (Li et al 2019) have been suggested.…”

“…Label-free deep learning methods have been created to reconstruct fluorescent images from brightfield images (Christiansen et al 2018, Ounkomol et al 2018, Nguyen et al 2021), but to our knowledge, the reconstruction of single-cell multichannel fluorescent images in IFC data has not been proposed. Label-free cytometry methods based on segmentation and unsupervised modeling (Nguyen et al 2021), weak supervision (Otesteanu et al 2021), cytometry by time of flight (CyTOF) (Hu et al 2020) and time-stretch microscopy (Li et al 2019) have been suggested. Previous methods have also utilized the Amnis IDEAS® software analyses, such as nuclear localization, or user generated cell masks (Lippeveld et al 2020), which allow for the segmenting of the cell from its surroundings based on pixel intensity.…”

Flexible and scalable control of T cell memory by a reversible epigenetic switch

“…3; Fig. S3; see Methods) 27 . Using this method, we tracked a total of 120 clonal lineages over 4 days and an average of 4.4 cell generations.…”

“…S3A-B), modified to enable segmentation of cells from brightfield movies. Importantly, to segment cells without additional fluorescent labels besides Tcf7-YFP, we first trained a convolutional neural network (CNN) with a U-net architecture 61 to predict fluorescence images of whole cells from brightfield images, using images of cell-trace violet labeled T cells as a training data set 27 . We trained separate CNNs for the images acquired in 96-well plates (Fig.…”