CellView: Interactive exploration of high dimensional single cell RNA-seq data

Abstract

Recent technological advances in single cell capture and nano-scale reactions have led to a major revolution in single cell transcriptomics 1,2,3 . Single cell datasets are analyzed using computational and statistical frameworks that enable feature (gene) selection, dimensionality reduction, clustering and differential gene expression. Multiple software packages exist that allow researchers well versed in computational analysis to perform this analysis [4][5][6] . However, identifying the exact parameters required for cell type identification is an iterative process greatly improved when informed by biology. In addition, interactive exploration of single cell datasets incorporating a biologist's knowledge greatly improves data interpretation, yet often such experts do not have big data handling skills.Advances in web application frameworks and visualization methods for dense datasets facilitate the development of interactive applications to allow easy and intuitive exploration of single cell data. Here, we introduce an R Shiny 7 web application, CellView, that allows knowledge-based and hypothesis-driven exploration of processed single cell transcriptomic data. The input into CellView is an R dataset (. . CC-BY-NC-ND 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint (which . http://dx.doi.org/10.1101/123810 doi: bioRxiv preprint first posted online with three pre-computed data frames containing expression, clustering, and gene symbol information.This file is agnostic of upstream computational approaches providing flexibility in algorithms used to calculate these data frames. This .Rds file can be shared with the end user, eliminating the need for hosting datasets, thereby decreasing the size of a virtual machine or cloud instance required to host and use CellView. Multiple tabs allow for easy access to the data and visualization of gene expression across and within clusters, aiding cell type identification.To illustrate the utility and power of CellView, we generated and analyzed single cell transcriptome data from peripheral blood mononuclear cells (PBMCs) using the 10X Genomics Chromium 8 . As defined by the CellRanger 8 pipeline, this data consisted of 6,554 single cells sequenced to 90.1% saturation with, on average, 824 genes and 2,077 molecules detected per cell. Dimensionality reduction using tSNE 9 was applied to genes selected by normalized dispersion, and with clustering by DBSCAN 10 . CellView automatically determines cluster numbers, updates the user interface, and renders a 3D scatter plot displaying cells clustered in tSNE space (Fig 1b) (Fig 1c), a marker of B-cells, and CD3D, a marker of T-cells (Fig 1d), provide representative views of the 'Explore' tab.The 'Co-expression' tab enables the generation of heatmaps to visualize expression of multiple genes either across all clusters, in the 'AllClusters' sub-menu, or on selected cells wi...

Cross-Species Single-Cell Analysis of Pancreatic Ductal Adenocarcinoma Reveals Antigen-Presenting Cancer-Associated Fibroblasts

“…Certain cell populations were subclustered by repeating the procedure described above. Interactive analysis was done using the CellView RShiny web application (83). The mouse RNA-seq data are available at the Gene Expression Omnibus under the accession number GSE129455.…”

Single-cell transcriptomic analysis of the lateral hypothalamic area reveals molecularly distinct populations of inhibitory and excitatory neurons

“…The analysis routine for the singlecell data is defined in the scRNA-seq analysis section and further details, if needed, provided upon request. Interactive analysis was done using the CellView RShiny web application 65 .…”

Single-cell transcriptomics–based MacSpectrum reveals macrophage activation signatures in diseases

“…FASTQ files were then aligned to mm10 mouse reference genome and transcriptome using the CellRanger v1.3 software pipeline with default parameters as reported previously (78); this demultiplexes the samples and generates a gene versus cell expression matrix based on the barcodes and assigns UMIs that enables determination of the individual cell from which the RNA molecule originated. Gene expression was normalized using CellView software (79). Briefly, the number of gene transcripts per cell was multiplied by the median of transcripts across all the cells, and then log2 transformed (following an addition of +1 pseudocount to prevent log error where the transcript count is 0; i.e., log2[0 + 1] = 0), resulting in normalized expression (NE) values.…”