Zubair Khalid

Virologist/Molecular Biologist | Veterinarian | Bioinformatician

Conventional & Molecular Virology • Vaccine Development • Computational Biology

Dr. Zubair Khalid is a veterinarian and virologist specializing in conventional and molecular virology, vaccine development, and computational biology. Dedicated to advancing animal health through innovative research and multi-omics approaches.

Dr. Zubair Khalid - Veterinarian, Virologist, and Vaccine Development Researcher specializing in Computational Biology, Multi-omics, Animal Health, and Infectious Disease Research

Category: Guides

Simultaneous Single Cell Atac Seq And Rna Seq

Simultaneous single cell ATAC seq and RNA seq is a multiomic method that profiles chromatin accessibility and gene expression from the same individual cell, enabling direct linkage of regulatory elements to transcription. This guide is for bench scientists, bioinformaticians, and core facility staff who plan to design, execute, or interpret such experiments. It provides a source bounded and practical framework covering core concepts, decision points, a step by step workflow, quality metrics, common mistakes, and limits of interpretation.

At a Glance

Aspect Key Information
What it measures Chromatin accessibility (ATAC) and gene expression (RNA) in the same single cell
Typical platforms 10x Genomics Multiome, sci CAR, droplet based or split pool barcoding
Main advantage Direct cis regulatory inference without computational imputation or matching
Key challenge Cell throughput vs. sequencing depth trade off, doublet rates, and data sparsity
Recommended read depth ATAC: 25,000 50,000 fragments per cell, RNA: 10,000 20,000 reads per cell
Primary quality metric Fraction of reads in peaks (FRiP) > 0.3, and RNA median transcripts per cell > 1,000

Core Concepts

Simultaneous single cell ATAC seq and RNA seq involves capturing both the open chromatin landscape and the transcriptome from the same nucleus. The most widely used commercial approach, the 10x Genomics Multiome assay, begins with a single nucleus suspension, then partitions nuclei into droplets containing barcoded beads. A reverse transcription reaction captures polyadenylated RNA, while a separate tagmentation step inserts sequencing adapters into accessible DNA. Both libraries are then sequenced independently, and the shared cell barcode links the two modalities. Bioconductor provides robust software for processing such multiomic data, including the Seurat and Signac packages, which handle joint normalization and downstream analysis.

The biological rationale is compelling. Chromatin accessibility is a proxy for regulatory potential, while gene expression reflects actual transcription. By measuring both in the same cell, researchers can infer which accessible regions are associated with which genes, build enhancer gene networks, and identify transcription factor binding motifs enriched in differentially accessible peaks. This approach bypasses the need for computational integration of separately profiled single cell RNA seq and ATAC seq datasets, which can suffer from confounding technical batch effects and require sophisticated matching algorithms. A benchmark study in Genome Biology highlights that while unpaired integration methods have improved, joint profiling remains the gold standard for direct cis regulatory inference. Benchmarking component choices for unpaired single cell RNA and epigenomic integration (source [8]).

Another core concept is library complexity. The ATAC library in a single cell experiment yields sparse coverage per cell, typically capturing only a few thousand unique fragments from the accessible genome. The RNA library captures a similar sparse transcriptome. Therefore, cell type identification is often driven by the transcriptome, with chromatin data used for downstream regulatory analysis. However, when both modalities converge on the same cell clusters, the confidence in cell state annotation is greatly increased. Recent work profiling peripheral blood with an optimized multiplexed multiome approach demonstrated that insulin driven asthma subtypes could be identified only when both chromatin and expression profiles were considered together. Profiling Peripheral Blood with an Optimized, Multiplexed, Single cell Multiome Approach Supports an Insulin driven Asthma Subtype (source [9]).

Decision Points

Before beginning a simultaneous single cell ATAC RNA experiment, you must make several key decisions that trade off cost, throughput, and data quality.

First, choose a platform. The 10x Genomics Multiome is the most widely used and well supported. Alternative methods include split pool barcoding approaches like sci CAR or combinatorial indexing, which can reduce per cell cost but increase library preparation complexity. Consider your lab's experience with microfluidics vs. plate based workflows. The Galaxy Training Network offers free tutorials for processing 10x Multiome data end to end. Galaxy Training Network (source [3]).

Second, decide on tissue dissociation and nuclei isolation. Simultaneous ATAC RNA requires intact nuclei, not whole cells, because cytoplasmic RNA contamination can reduce nuclear RNA capture. A gentle nuclei isolation protocol using lysis buffer with RNase inhibitors is critical. The quality of the nuclei suspension, measured by trypan blue exclusion and microscopy, directly impacts cell recovery and data quality.

Third, set your coverage goals. For typical mouse or human tissues, targeting 5,000 10,000 cells per sample is common. However, if you need to resolve rare subpopulations, you may need to increase cell loading, which raises doublet rates. A doublet rate of 5 10% is acceptable, but you must computationally doublet score and filter. The NCBI Sequence Read Archive contains many public Multiome datasets that can guide expectation of cell recovery and sequencing depth. NCBI Sequence Read Archive (source [5]).

Fourth, plan for sequencing. Paired end reads are required for ATAC libraries (to map fragment coordinates precisely), and single end or paired end is acceptable for RNA. Many facilities run ATAC at 50 bp paired end and RNA at 28 bp read1 plus 91 bp read2 for 10x. Balancing sequencing lanes between the two libraries is important. A common mistake is to over sequence the ATAC library while under sequencing the RNA library, leading to high ATAC saturation but insufficient RNA detection.

Workflow or Implementation Steps

The following practical workflow assumes you are using the 10x Genomics Multiome platform. Adaptation for other platforms is similar in principle but differs in barcoding and library prep.

  1. Nuclei preparation. Harvest tissue and generate a single cell suspension. Use a commercial or published nuclei isolation kit designed for single cell multiomics. Strain the suspension through a 40 micron filter, count nuclei, and evaluate quality. Aim for at least 90% intact nuclei and minimal debris. Centrifuge at 500 x g for 5 minutes at 4 degrees Celsius.

  2. Transposition and reverse transcription. Follow the 10x Genomics Next GEM Multiome protocol. Briefly, mix nuclei with transposition mix, incubate at 37 degrees Celsius for 1 hour. Then combine with reverse transcription master mix and load onto a chromium controller. The gel beads contain unique barcodes that will label both the ATAC and RNA fragments from the same nucleus.

  3. GEM incubation and cleanup. Perform reverse transcription in a thermal cycler. After reaction, break the emulsion and produce a single barcoded cDNA and ATAC fragment pool. Use SPRI bead cleanup to remove excess primers and small fragments.

  4. Library construction. The protocol involves separate PCR amplifications for the RNA and ATAC libraries, each with sample indices. The ATAC library typically uses a different set of primers that enrich for tagmented fragments. Run a Bioanalyzer or TapeStation to confirm library size distribution. ATAC libraries should show a nucleosomal ladder pattern, while RNA libraries should show a peak around 400 600 bp.

  5. Sequencing. Pool the two libraries in a ratio that reflects your desired read depth. For a typical experiment targeting 5,000 cells, sequence the ATAC library to 500 million paired end reads and the RNA library to 300 million reads. Use a NovaSeq 6000 or NextSeq 2000.

  6. Primary data processing. Use the 10x Genomics Cell Ranger ARC pipeline (or the open source cellranger arc). This pipeline demultiplexes, maps to the genome, counts fragments per cell for ATAC, and quantifies transcripts per cell for RNA. Generate the filtered feature barcode matrix.

  7. Secondary analysis. Read the output into R or Python. For Seurat, create a Seurat object with the RNA assay and add an ATAC assay using CreateAssayObject. Filter cells by RNA feature count (200 5000), mitochondrial percentage (<20%), and ATAC fragment count (500 30,000). Perform joint normalization using weighted nearest neighbor (WNN) analysis as implemented in Seurat v4 and later. This method learns a cell specific modality weight that balances the information from RNA and ATAC. The EMBL EBI Training materials provide a detailed walkthrough of downstream integration and plotting. EMBL EBI Training (source [2]).

  8. Regulatory inference. Call peaks on the merged ATAC data using MACS2 or the Signac package. Then compute per cell accessibility in peaks. Link peaks to genes using correlation of accessibility and expression across cells, or by using motif enrichment. Tools like Cicero or chromVAR can identify co accessible peaks and transcription factor activity.

Quality Checks

Quality control is essential because simultaneous multiomic experiments generate two sparse matrices that can be corrupted by technical artifacts. The following checks should be performed at each stage.

  • Nuclei quality check. Before loading, stain with DAPI and count. A high proportion of doublets or debris will ruin the experiment. Check for RNase contamination by leaving a small aliquot at room temperature for 30 minutes and re staining. If fluorescence drops, RNase is present.
  • Library QC. Run tapestation traces. The ATAC library should have a broad peak around 200 600 bp with a nucleosomal ladder. The RNA library should show a smooth distribution without adapter dimer peaks (<150 bp). If you see large peaks at 120 bp, you have adapter contamination and need to re clean.
  • Sequencing QC. Use FastQC and multiqc. Check per base sequence quality, GC content, and duplication levels. The ATAC library is expected to have higher duplication than RNA because of the lower complexity.
  • Cell barcode filtering. After Cell Ranger ARC, inspect the knee plot of barcode rank vs. total fragments. A clear knee indicates successful cell calling. If the knee is shallow, you may have high background or damaged nuclei. Filter cells with low fragment counts (below 500) and high mitochondrial RNA fraction (above 20% for RNA, but note that ATAC does not have mitochondrial reads).
  • Modality consistency. Plot the number of ATAC fragments vs. RNA UMI counts per cell. In healthy cells, there is a positive correlation. If you see many cells with high ATAC but very low RNA, those could be empty droplets with chromatin debris. Conversely, cells with high RNA but low ATAC might be RNA contaminants. Use of the doublet detection method scrublet or DoubletFinder on the RNA matrix can flag multiplets.
  • FRiP score. Compute fraction of reads in peaks for each cell. A typical FRiP score for single cell ATAC should be above 0.3. Lower values suggest poor tagmentation or high background. Filter cells with FRiP < 0.2.

Common Mistakes

Several recurring pitfalls can compromise a simultaneous single cell ATAC RNA experiment.

  • Using too many nuclei. Loading beyond the recommended recovery (e.g., >10,000 per channel) leads to high doublet rates and poor cell separation. Always target a recovery of 5,000 to 8,000 cells for a standard chip.
  • Insufficient RNase inhibition. Nuclei isolation often involves hypotonic lysis, which can activate endogenous RNases. Always include a potent RNase inhibitor in the lysis buffer and work quickly. If your RNA library shows low gene detection (median <500 genes per cell), suspect RNase activity.
  • Incorrect sequencing depth ratio. Many teams sequence the ATAC library much deeper than the RNA library, but both need sufficient saturation. A common mistake is to sequence ATAC to 100,000 fragments per cell while only achieving 5,000 RNA reads per cell. This imbalance limits the ability to identify cell states because the transcriptome defines cell identity. A rule of thumb is to target 25,000 50,000 ATAC fragments and 10,000 20,000 RNA reads per cell.
  • Ignoring ambient RNA and ATAC contamination. Empty droplets can contain ambient RNA from lysis. Cell Ranger ARC attempts to remove these via barcode filtering, but residual ambient RNA can still affect downstream clustering. Use cellbender or SoupX to correct ambient RNA in the RNA assay before joint analysis.
  • Forgetting batch effects. If processing multiple samples, do not pool raw data without attempting batch correction. A study on multiomic single cell perturbation screens found that batch effects can obscure relevant regulatory signatures if not addressed. Multiomic single cell perturbation screens reveal critical lncRNA regulators of senescence (source [10]). Use Harmony or Seurat integration for joint batch correction after WNN.

Limits of Interpretation

Simultaneous single cell ATAC RNA data is powerful, but its interpretation has inherent limits.

First, the data are extremely sparse. In any given cell, less than 5% of accessible regions are actually detected, and only about 10% of expressed genes are captured. Regulatory element gene linkages derived from correlation across cells aggregate information globally and may not hold at the single cell level. A new method, BiCLUM, uses bilateral contrastive learning to improve unpaired integration, but it still requires many cells to infer robust links. BiCLUM: Bilateral contrastive learning for unpaired single cell multi omics integration (source [11]).

Second, chromatin accessibility does not equal transcription factor binding. An open region may be bound by repressors or structural proteins, not directly by the transcription factor whose motif is present. Motif enrichment analysis suggests binding potential, but without direct protein DNA interaction data (e.g., CUT&Tag), you cannot confirm binding.

Third, the technique does not capture splicing isoforms, post transcriptional regulation, or protein abundance. Gene expression measured by polyA capture reflects mRNA level, not translation. For a complete picture, other modalities such as CITE seq (surface protein) or single cell proteomics would be needed.

Fourth, cell type annotation relies heavily on the transcriptome. If your RNA library is of low quality, you may not be able to resolve fine cell subtypes even if the ATAC data is high quality. Conversely, ATAC data can sometimes reveal regulatory differences between cells that look transcriptomically identical, but interpreting these differences requires validation.

Finally, the current cost and throughput limits prevent routine use for very large scale projects (e.g., >100,000 cells). While combinatorial indexing methods may reduce cost, they are less well standardized. The NCBI Bookshelf provides an overview of current technical limitations and emerging solutions in single cell multiomics. NCBI Bookshelf (source [1]).

Frequently Asked Questions

Can I use existing single cell RNA seq analysis tools on the RNA part of a Multiome experiment? Yes. The RNA library from a Multiome experiment is essentially a standard 3 prime single cell RNA seq library. Most tools, including Seurat, Scanpy, and Monocle, work without modification. However, you must account for the lower gene detection typical of single nucleus RNA seq compared to single cell RNA seq from whole cells. Some genes enriched in cytoplasm (e.g., ribosomal) may be underrepresented.

How do I decide whether to perform simultaneous profiling or separate assays with computational integration? Simultaneous profiling is recommended when you need to link regulatory elements to genes at single cell resolution, for example, to identify enhancer gene pairs in a rare population. If your goal is simply to map cell types and you already have separate single cell RNA and ATAC datasets, computational integration with methods like Seurat WNN or scVI may be sufficient. The benchmark study from Genome Biology (source [8]) offers a decision tree based on dataset size and desired resolution.

What is the recommended number of cells per sample for a typical multiome experiment? For most tissues, 5,000 to 8,000 cells after filtering is sufficient to identify major and minor cell types. If you are studying a very rare population (<1% of cells), you may need to profile 15,000 20,000 cells, but be mindful of higher doublet rates and increased sequencing cost. It is better to run two separate channels at moderate loading than one overloaded channel.

How do I handle samples with low viability after nuclei isolation? Low viability (<75% intact nuclei) often leads to high background and low quality data. If viability is low, you can try a density gradient centrifugation to enrich for intact nuclei. Alternatively, consider using a dead cell removal kit designed for nuclei. If the problem persists, it may be better to discard the sample and optimize the tissue dissociation protocol. Post sequencing, you can attempt to filter out cells with high mitochondrial RNA content and low fragment counts, but this will reduce your cell numbers.

References and Further Reading

Related Articles