Discovering Novel Cell Types across Heterogeneous Single-cell Experiments

Abstract

Although tremendous effort has been put into cell type annotation and classification, identification of previously uncharacterized cell types in heterogeneous single-cell RNA-seq data remains a challenge. Here we present MARS, a meta-learning approach for identifying and annotating known as well as novel cell types. MARS overcomes the heterogeneity of cell types by transferring latent cell representations across multiple datasets. MARS uses deep learning to learn a cell embedding function as well as a set of landmarks in the cell embedding space. The method annotates cells by probabilistically defining a cell type based on nearest landmarks in the embedding space. MARS has a unique ability to discover cell types that have never been seen before and annotate experiments that are yet unannotated. We apply MARS to a large aging cell atlas of 23 tissues covering the life span of a mouse. MARS accurately identifies cell types, even when it has never seen them before. Further, the method automatically generates interpretable names for novel cell types. Remarkably, MARS estimates meaningful cell-type-specific signatures of aging and visualizes them as trajectories reflecting temporal relationships of cells in a tissue.

Temporal evolution of single-cell transcriptomes ofDrosophilaolfactory projection neurons

“…Following this observation, we decided to use differentially expressed genes between 24h PN clusters for PN-type identification for all stages. We applied m et a -learned representations for s ingle cell data (MARS) for identifying and annotating cell types (Brbić et al, 2020). MARS learns to project cells using deep neural networks in the latent low-dimensional space in which cells group according to their cell types.…”

“…Unless otherwise specified, all data analysis was performed in Python using Scanpy (Wolf et al, 2018), Numpy, Scipy, Pandas, scikit-learn, and custom single-cell RNA-seq modules (Li et al, 2017; Brbić et al, 2020). Gene Ontology analysis were performed using Flymine (Lyne et al, 2007).…”

Single-cell transcriptomes of developing and adult olfactory receptor neurons inDrosophila

“…We used the recently developed method, meta-learned representations for single cell (MARS) to cluster 24h APF ORNs (Brbic et al, 2020). MARS groups cells according to their cell types using a deep neural network to learn a set of cell-type specific landmarks and embedding function to project cells into a latent low-dimensional space.…”

“…We used this set of genes to confirm ORN types at 42h APF, and to identify ORN types at 24h APF and adult. To cluster each dataset, we applied MARS (Brbic et al, 2020)-a metalearning approach for cell type discovery. MARS leverages annotations of previously annotated datasets to better separate cell types in an unannotated dataset.…”

“…All RNA-seq data analysis was performed in Python using NumPy, SciPy, seaborn, pandas, scikit-learn, Scanpy (Wolf et al, 2018) and custom scRNA-seq analysis modules (Li et al, 2017;Brbic et al, 2020).…”