Single Cell RNA Seq Protocol

Single-cell RNA sequencing (scRNA seq) is a technique that profiles gene expression in individual cells, revealing heterogeneity invisible to bulk assays. This guide is for experimental biologists who are planning their first scRNA seq experiment or evaluating whether the method fits their research question. It provides a practical, source bounded framework built on established training resources and recent benchmark studies. EMBL EBI Training offers foundational modules, while the Galaxy Training Network supplies workflow oriented tutorials. Use this guide to navigate the main decisions, execute a reliable protocol, and avoid common pitfalls.

At a Glance

Core Concepts and Decision Points

ScRNA seq answers questions about cell type composition, states, transitions, and rare populations. The central challenge is converting the transcriptome of a single cell into a sequencing library with minimal bias and technical noise. NCBI Bookshelf provides detailed chapters on experimental design, particularly the trade off between throughput and resolution. The first decision is capture method. Droplet based platforms (10x Genomics, Drop Seq) profile thousands of cells at low per cell cost but at the expense of coverage (most reads come from the 3’ end). Plate based methods (SMART Seq2, CEL Seq2) capture fewer cells but yield full length transcript coverage and better sensitivity for low expressed genes. Choose droplet based if your goal is to identify cell types in a complex tissue. Choose plate based if you need to detect splice variants or study small populations of interest (e.g., circulating tumor cells). PMC42426463 demonstrates magnetic microraft isolation for rare CTCs before scRNA seq, a plate compatible workflow. Another decision point: fresh vs. frozen tissue. Many protocols require viable single cells, but recent advances allow cryopreserved nuclei. PMC42437889 used scRNA seq on donkey skin and chestnut samples after snap freezing, showing that careful optimization can preserve transcriptomic profiles. However, freeze thaw cycles generally reduce RNA integrity and should be avoided when possible.

Workflow and Implementation Sequence

1. Sample Preparation

Dissociate solid tissue into a single cell suspension with minimal stress. Enzymatic digestion (collagenase, trypsin) must be optimized for each tissue to avoid RNA degradation or activation of stress genes. Filter cells through a 40 micron mesh and assess viability with Trypan blue or fluorescent dyes. Target viability above 85%. Lower viability increases background from dying cells and may clog microfluidic chips. For blood or liquid biopsies, enrich target cells (e.g., using FACS or magnetic beads). PMC42426462 details a workflow for pancreatic circulating tumor cells using fluorescence activated cell sorting prior to scRNA seq.

2. Cell Isolation and Barcoding

Load cells onto the chosen platform. In droplet systems, a microfluidic chip encapsulates single cells with barcoded beads and lysis buffer. Each bead carries a unique barcode that tags all transcripts from one cell. The key parameter is cell loading concentration: aiming for 500-1000 cells per microliter usually yields about 60-70% occupancy (single cells per droplet). Overloading increases doublets (two cells per droplet). Plate based methods require depositing single cells into wells, either by FACS, limiting dilution, or manual picking. Galaxy Training Network has a tutorial on quality control for droplet data, including doublet detection.

3. Reverse Transcription and Library Amplification

Barcoded mRNAs are reverse transcribed into cDNA. In 3’ counting protocols (e.g., 10x), only the region near the poly A tail is amplified, yielding a single read per transcript. In full length methods, the entire transcript is amplified. This step is prone to bias: low input RNA may require many PCR cycles, which can distort relative abundances and introduce duplicates. Use Unique Molecular Identifiers (UMIs) to mitigate PCR duplication bias. UMI counting is standard in droplet protocols. Bioconductor packages such as scater and scran provide robust methods for UMI based normalization. After PCR cleanup (bead based purification), check cDNA concentration and size distribution on a Bioanalyzer. A typical profile shows a broad peak from 200 bp to several kb.

4. Sequencing

Sequence libraries on an Illumina platform. For 3’ counting, typical reads are 28 bp for the cell barcode and UMI (read 1) and 91 bp for the transcript (read 2). Depth recommendations vary: 20,000-50,000 reads per cell is common for cell type identification, deeper sequencing (100,000+ reads per cell) may be needed for detection of rare transcripts or full length methods. NCBI Sequence Read Archive hosts thousands of scRNA seq datasets with metadata on sequencing depth and platform. It is worth consulting those to gauge realistic depths for your organism and tissue.

5. Data Analysis

Raw sequencing data are demultiplexed into a gene expression matrix. For 10x data, Cell Ranger (or free alternatives like STARsolo and kallisto bustools) aligns reads to a reference genome, assigns UMIs and barcodes, and outputs a count matrix. Quality metrics include: number of genes detected per cell (expect 500-3000 for mammalian cells), total UMI count, and percentage of mitochondrial reads (high mitochondrial reads correlate with damaged cells). Filter cells with fewer than 200 genes, more than 5000 genes (possible doublets), and >20% mitochondrial content. Normalization methods (e.g., scran deconvolution or sctransform) account for library size differences. Dimensionality reduction (PCA, UMAP) precedes clustering and differential expression. Galaxy Training Network provides a full pipeline for scRNA seq analysis, from fastq processing to marker gene identification.

Quality Checks and Common Mistakes

Viability over 85% is essential. Low viability leads to high ambient RNA that contaminates other cells’ expression profiles. A common mistake is using samples with viability below 80% and then relying on computational filtering to remove “dead cell” clusters. This wastes sequencing budget and may obscure rare populations. PMC42432033 applied scRNA seq to mouse lung tissue and emphasized the need for careful filtering to exclude damaged cells, particularly in atopic conditions where cell death is elevated.

Doublet rate. In droplet experiments, doublets typically range from 1% to 10% depending on loading concentration. Software tools (DoubletFinder, scrublet) can identify doublets computationally, but they are not perfect. The best prevention is optimization of cell loading. Overloading to maximize cell capture is a common mistake that increases doublet rate and confounds downstream analysis.

Library complexity. A low number of detected genes (e.g., < 500) may indicate poor RNA recovery or cell stress. Conversely, a very high number of genes per cell (e.g., > 8000) often indicates a doublet. Check the knee plot (UMIs per barcode) to separate cells from empty droplets. EMBL EBI Training warns against relying solely on automated cell calling, always inspect the distribution.

Batch effects. When processing multiple samples, each batch (chip, day, reagent lot) can introduce technical variation. Use experimental designs that balance conditions across batches. Data integration methods (Harmony, Seurat CCA, scVI) can correct batch effects, but they cannot recover missing cell types if batches are completely confounded. PMC42441075 benchmarks long read scRNA seq methods and underscores that batch effects remain a major challenge, especially when combining short read and long read data.

Limits and Interpretive Caution

ScRNA seq has inherent constraints. Dropout events (genes that are expressed but not detected) are common due to low RNA input. This means absence of evidence is not evidence of absence. Differential expression tests must be designed with zero inflation in mind. Additionally, the cell capture step cannot distinguish between truly low expression and stochastic noise. Biological interpretation should focus on robust differences (e.g., genes detected in >10% of cells in one cluster vs. another).

Another limit: scRNA seq captures a snapshot of expression, not dynamic processes. Trajectory inference can infer pseudotime, but it assumes continuous, gradual transitions. PMC42436560 used scRNA seq to study INHBA signaling in ovarian cancer, but the authors caution that functional validation was needed to confirm mechanistic roles, because scRNA seq alone does not prove causality.

Finally, cell type annotation relies on reference databases or prior knowledge. Unsupervised clustering may split a continuous population into arbitrary clusters. Always validate marker expression with independent methods (immunofluorescence, flow cytometry) on a subset of cells.

Frequently Asked Questions

What is the minimum number of cells needed for a scRNA seq experiment?

For droplet platforms, the minimum practical input is about 1,000 cells, though some protocols can work with as few as 100 cells if using a plate based method with manual isolation. Lower inputs increase the risk of capturing only a few hundred cells after filtering, which may be insufficient for detecting rare populations. PMC42426463 describes a workflow for as few as 50 CTCs using micro raft isolation and SMART Seq2.

How should I handle samples with low viability?

If viability is below 70%, consider enriching live cells by FACS or debris removal kits. Alternatively, use single nucleus RNA seq (snRNA seq) which is more tolerant of frozen or degraded tissue. However, snRNA seq captures nuclear RNAs and may miss cytoplasmic transcripts and low abundance mRNAs.

What is the optimal sequencing depth for scRNA seq?

For cell type identification using UMI based methods, 20,000 reads per cell is often sufficient. Deeper sequencing (50,000 reads per cell) improves detection of lowly expressed genes but yields diminishing returns due to the saturation of unique transcripts. For full length methods, aim for 1-2 million reads per cell to cover all isoforms. NCBI SRA records can guide you based on your organism.

Can I combine samples from different batches if I use a batch correction algorithm?

Yes, but batch correction works best when each batch contains multiple biological conditions. If all samples from condition A are in batch 1 and all from condition B are in batch 2, batch and condition are confounded, and correction may remove true biological differences. Plan your experiment so that replicates of each condition are spread across batches.

References and Further Reading

• NCBI Bookshelf: Single-Cell RNA Sequencing Experimental Design
• EMBL EBI Training: Single Cell RNA Seq Analysis
• Galaxy Training Network: ScRNA Seq Workflow
• Bioconductor: Single Cell Package Documentation
• NCBI Sequence Read Archive: Public scRNA Seq Data
• Benchmark of long-read scRNA seq methods (2024)
• Multiomic integration in donkey tissues (2024)
• INHBA signaling and scRNA seq in ovarian cancer (2024)
• scRNA seq in mouse lung model (2024)
• Microraft isolation of CTCs for scRNA seq (2024)

Related Articles

• Somatic Cell
• Phases In Cell Cycle
• Endoplasmic Reticulum Cell Function
• Stages Of Cell Cycles
• Pcr Test