Characterization of cell type, while once dominated by pathological description, has over the past decade shifted towards a more quantitative and molecular approach. As such, molecular measurements in single cells have emerged as the centerpiece of the current paradigm of mechanistic biological investigation (Trapnell, 2015). Technological advancements have enabled researchers to measure all aspects of the central dogma of molecular biology at the single-cell level (Stuart and Satija, 2019). Single-cell RNA sequencing (scRNA-seq), a technique that profiles the relative expression of genes in individual cells and single-cell Assay for Transposase Accessible Chromatin using sequencing (scATAC-seq), a technique that surveys genome-wide chromatin accessibility are the most well-established and widely-used of these methods (Buenrostro et al., 2015; Lähnemann et al., 2020). In fact, combined scRNA-seq + scATAC-seq assays are now routine (Perkel, 2021). Additionally, assays to profile DNA methylation (Luo et al., 2018) or protein levels are now maturing and becoming more widely-accessible (Specht et al., 2019; Labib and Kelley, 2020). Most recently, combinations of various data modalities can now routinely be collected in parallel from the same cell (Chen S. et al., 2019; Zhu et al., 2019; Ma et al., 2020; Xing et al., 2020; Swanson et al., 2021).

scRNA-seq experiments generate on the order of millions of sequencing reads that sample the relative expression of approximately 20,000–30,000 transcribed features (e.g., genes) in each cell of the sample. Normalized read counts for each feature can then be compared to discern differences between cells. scATAC-seq samples comprise a larger feature space wherein cells are characterized by the genomic coordinates of chromatin accessible regions and sequence-features derived from these regions (e.g. transcription factor motifs, k-mer frequencies, etc.). Initially performed in dozens to hundreds of cells, these experiments are now performed on the order of millions of cells. With a high dimensional feature space as a result of thousands of features being considered for each cell and large (in cell number) experiments, analysis methods for this data have been required to advance concurrently with the development of these technologies (Chen et al., 2019c; Tian et al., 2019).

With the exception of proof-of-concept methods still too nascent to be widely applied (Chen et al., 2021), omics measurements of single-cells are generally destructive, preventing measurement of a cell at more than a single time point. As a result, most single-cell measurements for studying dynamic processes are of a “snapshot” nature, imposing inherent limitations on the study of such processes from this data (Weinreb et al., 2018). In light of this, transcription rates can be informative of ongoing processes in cells. The recent advent of RNA velocity quantifies and models the ratios of spliced and unspliced RNA (mRNA and pre-mRNA, respectively) such that they indicate the temporal derivative of gene expression patterns and thereby reflect dynamic cellular processes, allowing predictions of past and future cell states (La Manno et al., 2018).

Among others, PCA, t-SNE, and UMAP are dimensional reduction methods that have become common choices for enabling the visualization of high-dimensional single-cell datasets. Dimensionally reduced datasets are plotted such that similar cells cluster together and those with highly differing features are likewise clustered apart. In addition to the visualization and clustering of cells, trajectory inference methods have been proposed to learn a latent topological structure to reconstruct the putative time-ordering (pseudotime) by which cells may progress along a dynamic biological process (Saelens et al., 2018). As single-cell technologies have advanced, techniques to cluster and organize cells based on single-cell assays have advanced alongside them, allowing key insights toward cell type and state characterization. Combined with RNA velocity information, trajectory inference can offer key insights on dynamical changes to cell states. Once in press however, representation of these dimensionally-reduced visualizations is limited to just two or three dimensions. Even using three-dimensional plots from published studies, one cannot dynamically adjust or rotate the visualization to better understand the data from another angle. In addition, cells are typically annotated by features (e.g. time points, cell type or clusters) to investigate stratification along an axis of some biological process. To change the annotations presented in publication, one must often reprocess the raw data, which is time- and skill-intensive, highlighting the need for more dynamical visualization tools. While such current data representations are often limited and static, single-cell omic datasets are information-rich and, in many cases, important biological heterogeneity cannot be easily investigated or visualized outside the scope of the original publication, without spending considerable cost and time to reanalyze the datasets from scratch.

VR visualization methods for single cell data have been recently proposed (Yang et al., 2018; Legetth et al., 2019; Bressan et al., 2021). However, these methods require either expensive hardware or specific data inputs that mandate intermediate to advanced computational skills. Thus, tools and clear protocols are required to enable researchers, especially those who are not able to efficiently reprocess the raw data, to explore the richness of published datasets (or their own unpublished data) through a simple, easy and affordable VR platform. Importantly, this platform must be flexible enough to accept all types of omics data from established and emerging technologies and processing tools currently employed by the single-cell community.

At the time of this writing, three non-peer-reviewed methods employing VR technology that produce two- and three-dimensional visualizations of single-cell data have recently been reported. CellexalVR enables the visualization of standard scRNA-seq data though requires users to preprocess their data through scripting (Legetth et al., 2019). Unfortunately, this tool also requires expensive and dedicated VR hardware to operate. Another recent method for visualizing single-cell data in VR is Theia (Bressan et al., 2021), which has been designed with a focus on the exploration of spatial datasets for both RNA and protein measurements. Similar to CellexalVR, expensive computing power and VR hardware required to use Theia creates a barrier to entry. An alternative to these high-performance methods for VR visualization of single-cell data is starmap (Yang et al., 2018), which allows the use of inexpensive cardboard visor hardware. However, starmap lacks the advanced portability of outputs from commonly-used scRNA-seq analysis tools and limits cell annotation to clustering results of transcriptomic data. Of note, there are currently no peer-reviewed tools available for the visualization of single-cell data in VR illustrating the novelty in this area of research. To overcome the limitations of these existing methods as well as build on their qualities and initial progress, we present singlecellVR, an interactive web application, which implements a flexible, innovative visualization for various modalities of single-cell data built on VR technology. singlecellVR supports clustering, trajectory inference and abstract graph analysis for transcriptomic as well as epigenomic and proteomic single cell data. Importantly, singlecellVR supports visualization of cell dynamics as described by RNA velocity, a recent milestone in the sequence-based analysis of single cells (La Manno et al., 2018; Bergen et al., 2020). singlecellVR is a browser-contained, free, and open-access tool. Notably, we have developed a one-command conversion tool, scvr to directly prepare the results of commonly-used single-cell analysis tools for visualization using singlecellVR.

The amount of publicly available scRNA-seq data has exploded in recent years. With new assays to capture chromatin accessibility, DNA methylation and protein levels in single cells, we predict a second wave of dataset generation. Each of these datasets is extremely high-dimensional and thus, rich with latent information about a given biological sample. Ideally, biologists would be able to explore this treasure-trove of data from any angle and make hypotheses assisted by in silico analysis at little to no time cost. Often however, experimental biologists lack the advanced computational skills and/or time required to reprocess and reanalyze raw data from published experiments to gain an understanding of the data from their desired angle of interest. Additionally, biologists who wish to thoroughly explore data prior to publication may rely on a computational specialist who is less connected to the biological problem of interest, introducing a disconnect in hypothesis-driven experimental turnover.

While once primarily reserved for entertainment, VR has found utility in both industrial and academic applications. In this manuscript we present a protocol for visualizing single-cell data in VR. This protocol is based on singlecellVR, a VR-based visualization platform for single cell data and discusses its innovations and differences with existing methods. Importantly, we provide a simple mechanism to prepare results from commonly-used single-cell analysis tools for VR visualization with a single command to considerably increase accessibility (see Section 5). With this added utility, we seek to empower non-computational biologists to explore their data and employ rapid hypothesis testing that could not be made from the traditional static representations typical of communication in a scientific report on paper or a computer screen.

We anticipate that VR will become increasingly useful as a research and education tool and that the construction of software libraries will aid such advancements. VR has also recently found application in other sources of biological data, including single-neuron morphological imaging data (Wang et al., 2019), three-dimensional confocal microscopy data for fluorescent molecule localization (i.e., fluorophore-tagged proteins) within cells (Stefani et al., 2018), and three-dimensional single-molecule localization super-resolution microscopy (Spark et al., 2020). Our scalable and flexible VR visualization framework is not limited to scRNA-seq and it can be also easily adapted to other single-cell assays and tools that already support epigenomic data and/or single-cell proteomic data (EpiScanpy (Danese et al., 2019), Seurat (Stuart et al., 2019), and STREAM (Chen et al., 2019a)). Finally, we extend our framework to computational methods that derive the RNA velocity of single cells for visualization in VR (La Manno et al., 2018; Bergen et al., 2020). With the recent advances in spatially-resolved transcriptomics (Welch et al., 2019) and corresponding analysis methods (Hao et al., 2021; Miller et al., 2021), visualization of such data has already been extended to a VR framework (Bressan et al., 2021). We believe this new sort of three-dimensional VR will also become especially useful once made available to the general research community via inexpensive hardware and facile data preprocessing and preparation for VR visualization. As software to analyze single cells reach their maturity, one could imagine the incorporation of such visualizations into more clinically translatable settings, such as medical devices.

This manuscript presents singlecellVR, a scalable web platform for the VR visualization of single-cell data and its associated preprocessing software, scvr, which streamlines the results of commonly used single-cell workflows for visualization in VR. singlecellVR enables any researcher to easily visualize single-cell data in VR. The platform is user-friendly, requires no advanced technical skills or dedicated hardware. Importantly, we have curated and preprocessed several recent single-cell datasets from key studies across various modalities of data generation and analysis approaches, providing the scientific community with an important resource from which they may readily explore and extract biological insight.