Interactive single-cell data analysis using Cellar

Abstract

Cell type assignment is a major challenge for all types of high throughput single cell data. In many cases such assignment requires the repeated manual use of external and complementary data sources. To improve the ability to uniformly assign cell types across large consortia, platforms and modalities, we developed Cellar, a software tool that provides interactive support to all the different steps involved in the assignment and dataset comparison process. We discuss the different methods implemented by Cellar, how these can be used with different data types, how to combine complementary data types and how to analyze and visualize spatial data. We demonstrate the advantages of Cellar by using it to annotate several HuBMAP datasets from multi-omics single-cell sequencing and spatial proteomics studies. Cellar is open-source and includes several annotated HuBMAP datasets.

Fig. 1. Cellar’s workflow.

a–c Preprocessing (optional). Cellar can filter cells based on the number of expressed genes, and genes which are rarely expressed. Next the input is normalized. d, e Dimensionality reduction and visualization. Several methods for dimensionality reduction are implemented as part of Cellar. The reduced data is then visualized by running another (possibly the same) dimensionality reduction method. f–i Clustering. Cellar supports several unsupervised and semi-supervised clustering methods. It also implements supervised label transfer methods. j–l Cell-type assignment. Cellar enables the use of several functional annotation databases for the assignment of cell types.

Fig. 2. CODEX data analysis in Cellar.

(ID: 19-003 lymph node R2) (a) UMAP visual representation of a lymph node CODEX dataset with 46,840 cells, clustered via Leiden. b Projection of the assignments on the spatial CODEX image that can be visualized side by side in Cellar. Cluster assignments were copied from (a). Not all clusters could be assigned to unique cell types given that only a few ten protein levels are measured, though several have been assigned based on differential gene analysis in Cellar. The B-Cell clusters are surrounded by T-cells and other cells types in the lymph. The B cell clusters also contain a subset of proliferating cells.

Fig. 3. SNARE-seq data analysis in Cellar.

(IDs: kidney SNARE ATAC/RNA 20201005) (a) UMAP plot of the chromatin modality for the kidney SNARE-seq dataset with 31,758 cells. First, we obtain a cell-by-cistopic matrix by running cisTopic which is then used to define clusters via Leiden clustering. b Corresponding UMAP plot of the expression matrix with cluster assignments copied from a. Cellar’s dual mode allows a cell ID based label transfer from one modality to the other.