The biological processes within each organism are carried out by complex networks of cells and molecules. The physiology of each tissue depends on proper signaling and interplay among the cells that comprise the structure, and the function of each cell is in turn dictated by its gene regulatory networks, which consist of cell-type-specific transcription factors and cis-regulatory sequences. The understanding of physiology and of disease pathogenesis will benefit greatly from systematic analysis of the biological networks within each cell, tissue and organ system in the body. Although genome sequencing has made it possible to assemble the part lists of biological networks, we envision that technological advances in single-cell multimodal omics will significantly accelerate the dissection of network connectivity and operational principles.

Virtually all cell types in the body share an identical genome. The identity and function of different cell types are determined by the cell’s epigenome—the collection of covalent modifications to histones and DNA that changes during development and influences the transcriptional output of the genome in each cell. Analysis of the epigenome, transcriptome and proteome together in different cells would allow the characterization of cell identity and state and the dissection of cell-type-specific gene regulatory networks. Recent advances in single-cell genomics technologies, which allow the profiling of genome, transcriptome and epigenome individually in single cells either one at a time or in a highly parallel fashion1, have greatly improved our ability to interrogate the biological networks in heterogeneous tissues. However, analysis of individual modalities in single cells still runs the risk of providing only a partial picture of the cells’ complex gene regulatory networks. Measurements of two or more modalities simultaneously, made possible by single-cell multimodal omics techniques, would enable a more complete understanding of the cellular and molecular circuitry regulating the function of each tissue and organ.

In this Commentary, we divide the single-cell multimodal omics approaches into two categories. The first category includes assays that measure multiple modalities one cell at a time, either in individual test tubes or in microwells of plates. These assays aim to achieve comprehensive profiling via multiple modalities in each cell and are limited by the throughput and relatively high cost of each assay (Fig. 1, methods indicated in blue). The second category of methods can process thousands to millions of cells together, through the use of droplet platforms or combinatorial DNA barcoding strategies, thus achieving high scalability and cost effectiveness (Fig. 1, methods indicated in brown).

Fig. 1: Methods for single-cell multimodal omics analysis.
figure 1

Methods have been developed to simultaneously profile epigenetic features, DNA sequences, gene expression perturbation and cell surface proteins in single cells. Single-cell multimodal omics methods designed for the analysis of one cell at a time are highlighted in blue, and highly scalable methods for analysis are highlighted in brown.

Full size image

Multimodal omics analysis: one cell at a time

How sequence variation affects phenotypic traits is a fundamental question of biology. To connect sequence variation to molecular phenotypes such as RNA expression within individual cells, DR-seq (gDNA-mRNA sequencing)2 and G&T-seq (genome and transcriptome sequencing)3 were developed for simultaneous genome sequencing and transcriptome profiling from the same cells: DR-seq uses specific primers for mRNA amplification to enrich RNA from DNA reads, and G&T-seq physically separates mRNA from genomic DNA with beads for each measurement. Obtaining the joint profiles of genome and transcriptome from the same single cells helped to identify the causative genetic variations for variable cell-to-cell expression in tumors. By combining scBS-seq (single-cell bisulfite sequencing, a means to detect methylated cytosine residues) with joint genome and transcriptome analysis, scM&T-seq (single-cell methylome and transcriptome sequencing)4 and the related scMT-seq5, scTrio-seq (single-cell triple omics sequencing)6 and snmCT-seq (single-nucleus methylcytosine and transcriptome sequencing)7 analyze the DNA methylome together with transcriptome in individual cells. Integrated DNA methylome and transcriptome analysis established the connection between heterogeneously methylated elements with variable gene expression in mouse embryonic stem cells4. The ability to probe the epigenomic–transcriptomic associations in single cells provides the potential to uncover functional regulators of stem cell maintenance as well as cell state transition and differentiation.

To dissect the crosstalk between different epigenetic layers, scCOOL-seq (single-cell chromatin overall omic-scale landscape sequencing)8 and scNOMe-seq (single-cell nucleosome occupancy and methylome sequencing)9 profile nucleosome occupancy and methylome in single cells. The joint epigenetic profiles provided insight into the relationship between dynamic chromatin states and DNA methylation during mouse early embryonic development8. This multimodal epigenetic analysis has been further combined with transcriptomic analysis in snNMT-seq (single-cell nucleosome, methylation and transcription sequencing)10 and scNOMeRe-seq (single-cell nucleosome occupancy, methylome and RNA expression sequencing)11, leading to tri-omics profiles of mouse early embryos and embryonic stem cells that revealed dynamic coupling of different molecular modalities during development and differentiation.

Mapping of accessible chromatin in single cells is a powerful approach for dissecting tissue heterogeneity and delineating candidate regulatory genomic sequences in each constituent cell type1. scCAT-seq (single-cell chromatin accessibility and transcriptome sequencing)12 and combined ATAC-RNA-seq (combined assay for transposase-accessible chromatin using sequencing and RNA sequencing)13 allow simultaneous profiling of open chromatin and gene expression by physically separating the genomic DNA and mRNA, similar to scTrio-seq5 and G&T-seq3, and for parallel open-chromatin and RNA sequencing from the same single cells. T-ATAC-seq (transcript-indexed ATAC-seq) combines single-cell ATAC-seq with expression analysis of T cell receptor (TCR) genes using the C1 microfluidics platform (Fluidigm) to link T cell specificity and open chromatin in individual cells14.

Higher-order chromatin architecture is another critical layer of gene regulation during development and in disease pathogenesis. By combining the capture of three-dimensional genome structure with profiling of the DNA methylome, scMethyl-HiC (single-cell Methyl-HiC)15 and snm3C-seq (single-nucleus methyl chromatin conformation capture sequencing)16 revealed the heterogenous nature of both chromatin conformation and DNA methylation in cell populations and in brain tissues, delineated the cell-type-specific chromatin architecture with annotation from methylome data and suggested the association between these two epigenetic layers. Recently, one high-resolution microscopy-based approach, known as ORCA (optical reconstruction of chromatin architecture), was developed in combination with fluorescence in situ hybridization targeting RNA (RNA-FISH) for parallel visualizing of DNA folding and gene expression in single-cells17, connecting the cell-type-specific chromatin architecture with gene expression.

Multimodal omics analysis: achieving scalability

Multimodal omics analysis of one cell at a time allows in-depth profiling of individual cells with relatively high genomic coverage and quantitative precision. However, the ability to profile only tens to hundreds of cells per experiment and the high cost per cell are quite limiting, especially in cases (such as the brain) where the heterogeneity of tissue samples is complex, necessitating the profiling of a large number of cells to gain an accurate view of the cellular composition in normal and disease conditions. To address this limitation, several approaches have been developed to enable high-throughput single-cell analyses of different modalities from thousands or more cells at a time.

Using droplet-based approaches, Perturb-seq18,19 and CRISP-seq20 pair CRISPR-based transcriptional interference with high-throughput single-cell RNA sequencing (scRNA-seq), with each single cell acting as an individual reporter for responses to genetic perturbation. These rich phenotypic data were used to reconstruct molecular circuits, with the ability to separate perturbation responses from confounding effects. Similarly, Perturb-ATAC-seq21, based on the C1 microfluidics platform, measures changes in chromatin states during CRISPR perturbations. The intercellular chromatin states variation was used to construct dynamic regulatory circuits and uncovered regulators acting during cell fate decisions. CITE-seq (cellular indexing of transcriptomes and epitopes by sequencing)22 and REAP-seq (RNA expression and protein sequencing)23 convert the abundances of different cell surface proteins to DNA barcodes, that can be captured together with mRNA by droplet-based scRNA-seq assays. Linking the view of proteins with transcriptome measurement enhances analysis of immune cell phenotypes and integrates the unbiased measurement of proteins into the molecular model for dynamic cellular responses. By expanding antibody barcoding to additional features, ECCITE-seq (expanded CRISPR-compatible CITE-seq)24 enables the multiplexed detection of the transcriptome together with cell surface proteins, clonotypes and CRIPSR perturbations in the same single cells. The combined TCR clonotype and surface protein information allowed fine definition of cellular phenotypes, and the low drop-out rate of protein detection also reduced the number of cells needed for a given CRISPR perturbation, providing richer phenotypic readouts each time.

Combinatorial single-cell indexing platforms, based on split-and-pool DNA-barcoding strategies, offer another highly scalable approach comparable to the droplet systems, but without the need for specific instruments. One method, known as single-cell combinatorial indexing (sci) and involving two rounds of splitting and pooling, was used to jointly profile the chromatin accessibility and gene expression from thousands of single cells (via sci-CAR, or single-cell combinatorial indexing joint profiling of chromatin accessibility and mRNA25). A similar combinatorial barcoding strategy involving multiple rounds of ligation-based combinatorial barcoding, Paired-seq (parallel analysis of individual cells for RNA expression and DNA accessibility by sequencing)26, allowed joint profiling of transcriptome and open chromatin in millions of single nuclei. Such single-cell multimodal omics analysis could also be performed with a droplet-based platform known as SNARE-seq (single-nucleus chromatin accessibility and mRNA expression sequencing)27. The combined open chromatin–transcription profiles of a large number of single cells could connect regulatory inputs with expression outputs and annotate cells with transcriptional states. The multilayer regulatory programs profiled by these methods will also, in our view, provide excellent chance for identifying both cis- and trans-functional regulators during development.

Challenges and opportunities

Although great progress has been made with single-cell multimodal omics technologies, many challenges remain. Multimodal omics analysis of one cell at a time provides comprehensive molecular profiles, making this ideal for analyzing biological samples consisting of a limited number of cells (for example, mammalian early embryos). However, due to the relatively limited throughput and often high per-cell costs, these methods are costly for large-scale analysis of complex heterogeneous samples and for identifying rare cell types within a tissue (for example, investigation of tissue-resident progenitor cells or organismal-level cell atlas analysis). On the other hand, one key limitation of the high-throughput single-cell multimodal omics assays is data sparsity. The coverage of epigenome and transcriptome of individual cells with the current approaches is still low: it is difficult to distinguish technical noise from cell-to-cell variability. Although optimization of experimental procedures may in the future narrow the gaps, fundamentally new biochemistry or strategies may be necessary to completely overcome this limitation (Fig. 2a).

Fig. 2: Challenges and opportunities in single-cell multimodal omics.
figure 2

a, Optimization of single-cell multimodal omics methods is needed to improve detection sensitivity and coverage. b, Co-assays of histone marks or transcription-factor binding and gene expression will refine the understanding of gene regulatory networks. c, Joint analysis of transcripts and proteins in single cells will reveal the dynamic relationships of transcripts and protein abundance. d, Extension of the spatial transcriptomics toolbox to epigenome analysis will facilitate the dissection of molecular and cellular networks in complex tissues.

Full size image

Also missing from the current single-cell multimodal omics approaches is an important molecular layer of regulatory control provided by various histone modifications and transcription factors (Fig. 2b). Using Tn5 transposase fused to staphylococcal protein A to bind and tag the genomic DNA associated with antibody-targeted regions enabled single-cell profiling of DNA binding proteins and histone modifications28. This approach could in principle be combined with sci-CAR25, SNARE-seq27 and Paired-seq26 to permit parallel measurement of histone modifications or transcription-factor occupancy and transcriptome. Additionally, we envision potentially fascinating results from linking perturbations of specific transcription factors to chromatin accessibility as well as gene expression in the same single cells.

It will also be crucial to perform combined analysis of single-cell transcriptomes and proteomes (Fig. 2c). This will shed critical light on the relationship between transcriptome and protein abundance and on the dynamic relationship in different cell types or states and under different developmental and/or disease conditions. RNA and proteins are very different molecular modalities, and converting them into a uniform form that can be captured together in an “omics-wide” manner is a substantial technical challenge at this stage. Probing the two-dimensional profiles in parallel would require high-throughput single-cell proteomics methods, which are not yet established.

Cellular states (epigenetic and transcriptomic) and spatial distribution are key determinants of cellular identity in mammalian tissues: cells with defined molecular profiles require strictly regulated spatial positioning for normal function. An important and fascinating area of further development is combining single-cell multimodal omics methods with spatial information, using approaches such as Slide-seq29 that convert spatial information into DNA barcodes, or imaging-based approaches30,31 that naturally preserve the spatial information of each cell. Extension of the in situ profiling toolbox to probe DNA methylation and chromatin accessibility would be a critical step forward, and we believe that such approaches will help to decipher the organizing principle of primary tissues in the body (Fig. 2d).

In summary, single-cell multimodal omics methods have greatly expanded our toolkit for delineating the complex molecular and cellular networks operating in diverse biological systems. New technological advances that increase coverage, add additional layers of information or enhance the scalability of assays would further empower investigators in pursuit of a comprehensive understanding of the cells.