Rna Sequencing Protocol

RNA sequencing (RNA seq) is a high throughput method for capturing and quantifying the transcriptome, the complete set of RNA transcripts in a cell or tissue. This guide is intended for molecular biologists, bioinformaticians, and students who need a practical, source bounded framework to design, execute, and interpret an RNA seq experiment. The protocol outlined here draws on authoritative training resources and peer reviewed methods to help you avoid common pitfalls and produce reproducible results. For foundational background, consult the NCBI Bookshelf collection of molecular biology references NCBI Bookshelf. For structured learning about sequencing analysis, the EMBL EBI training materials provide hands on modules EMBL EBI Training.

At a Glance

Core Concepts

RNA seq measures the abundance of RNA molecules by converting them into a complementary DNA (cDNA) library, sequencing the fragments, and mapping the reads to a reference genome or transcriptome. The fundamental steps include RNA extraction, library preparation, sequencing, and computational analysis. The Galaxy Training Network offers a complete tutorial on the RNA seq workflow, from raw reads to differentially expressed genes Galaxy Training Network. An important concept is that RNA seq provides relative abundance, not absolute molecule counts, unless spike ins are used. The Bioconductor project provides extensive documentation on statistical methods for RNA seq, including normalization and differential expression Bioconductor. You must account for biases like GC content, transcript length, and sequencing depth when interpreting results.

Decision Points

Before starting, you need to make several choices that affect the protocol.

Stranded vs unstranded library. Stranded protocols preserve the orientation of the original RNA molecule. This improves mapping accuracy for overlapping transcripts and is essential for applications like antisense RNA detection and 3 prime end sequencing. Unstranded libraries are simpler but lose strand information.

PolyA selection vs ribosomal RNA depletion. PolyA selection enriches for messenger RNA (mRNA) by binding the polyA tail. It is suitable for eukaryotic studies focused on protein coding genes. Ribosomal RNA depletion removes abundant rRNA and retains non coding RNAs, long non coding RNAs, and nascent transcripts. For prokaryotes, rRNA depletion is typically used because most bacterial mRNA lacks polyA tails.

Single end vs paired end reads. Single end reads are cheaper and sufficient for gene level quantification. Paired end reads improve mapping to repetitive regions, enable detection of structural variants, and are required for isoform level analysis. For most differential expression experiments, 50 to 75 base pair single end reads are adequate, but longer reads or paired ends provide added confidence.

Sequencing depth. Depth is the number of reads per sample. For gene level quantification, 10 to 30 million reads per sample is typical. For detecting rare transcripts or performing alternative splicing analysis, 50 million or more reads may be needed. The Sequence Read Archive (SRA) at NCBI contains metadata from thousands of RNA seq experiments that can guide depth planning NCBI Sequence Read Archive. A recent benchmark study on nanopore long read RNA seq highlights the trade offs between depth and technology choice for single cell and spatial applications A systematic benchmark of bioinformatics methods for single-cell and spatial RNA-seq nanopore long reads data, NAR Genom Bioinform 2025.

Biological replicates. A minimum of three biological replicates per condition is strongly recommended to power statistical tests. Fewer replicates may still detect large fold changes but increase false discovery rates.

Workflow or Implementation Steps

The following sequence provides a reproducible path from sample to results.

Step 1: RNA extraction and quality assessment. Extract total RNA using a method suited to your sample type, such as TRIzol, column based kits, or magnetic beads. Evaluate RNA integrity using an Agilent Bioanalyzer or similar platform. Aim for an RNA Integrity Number (RIN) above 7 for most mammalian tissues. Degraded RNA introduces 3 prime bias and reduces the number of detectable transcripts.

Step 2: Library preparation. Choose a commercial or core facility protocol for cDNA synthesis and adapter ligation. Stranded libraries are now standard. Include unique molecular identifiers (UMIs) if you plan to count absolute molecules and remove PCR duplicates. Perform a final PCR amplification with as few cycles as possible to avoid over representation. Library quality is checked by fragment length distribution (typically 200 to 600 bp) and concentration.

Step 3: Sequencing. Pool libraries equimolarly and sequence on an Illumina platform (e.g., NovaSeq, NextSeq) or other high throughput system. Ensure the read length covers the fragment insert. For standard differential expression, 1x75 bp single end on a lane of a flow cell often suffices. For isoform discovery, use 2x150 bp paired end.

Step 4: Preprocessing of raw reads. Run fastqc to assess per base quality, GC content, adapter contamination, and duplication rates. Trim low quality bases and adapters using tools like Cutadapt or Trimmomatic. Re run fastqc on trimmed reads to confirm improvement. This step is critical because adapter residues misalign and degrade quantification accuracy.

Step 5: Alignment or quantification. Map reads to a reference genome using a splice aware aligner such as STAR or HISAT2. Alternatively, use a pseudoalignment tool like Salmon or kallisto for faster transcript level quantification. Both approaches require a high quality annotation file (GTF/GFF). For human or mouse data, use the latest GENCODE release. The Galaxy Training Network provides exact command line parameters for these tools Galaxy Training Network.

Step 6: Gene level count matrix. If you used an aligner, count reads overlapping genes with featureCounts or htseq count. For pseudo alignment, tools produce a matrix directly. Ensure you set strandedness correctly in the counting step, mismatching strandedness eliminates most counts.

Step 7: Normalization and differential expression. Import counts into R/Bioconductor packages such as DESeq2, edgeR, or limma voom. Build a model matrix with condition and batch variables. Perform size factor normalization (DESeq2) or trimmed mean of M values (TMM, edgeR). Fit a negative binomial model and extract differentially expressed genes using a false discovery rate cutoff of 0.05.

Step 8: Post analysis and validation. Annotate results with gene symbols, fold changes, and p values. Perform gene set enrichment analysis using tools like clusterProfiler. Validate a handful of top hits using RT qPCR on the same RNA samples. For example, a dual RNA seq study identified virulence factors and host immune responses by combining transcriptomic data with functional validation in crustacean hemocytes Dual RNA-seq Identifies Virulence Factors in Mesanophrys sp. and Immune Defenses in Portunus trituberculatus Hemocytes, Acta Parasitol 2025.

Quality Checks

Implement these checks throughout the protocol.

RNA quality. Run a gel or Bioanalyzer trace before library prep. A low RIN indicates degradation, reprocess the sample if possible.

Library quantification. Use a fluorometric method (Qubit) and a fragment analyzer to confirm correct size distribution. A broad peak with a shoulder at primer dimer size suggests cleanup problems.

Sequencing metrics. Monitor cluster density, percent PF reads, and Q30 scores. For a typical run, Q30 > 80% is acceptable.

Alignment statistics. Report uniquely mapped read percentage. Values below 60% may indicate poor RNA quality, ribosomal contamination, or incorrect reference. Check the percentage of reads assigned to features, a low assignment rate ( < 40%) suggests annotation mismatch or strand setting error.

Biological variation. Use principal component analysis (PCA) on normalized counts. The first two components should separate your primary groups. Excessive dispersion within replicates may indicate a technical artifact.

Differential expression sanity. Confirm that known positive controls (e.g., sex specific genes if comparing male vs female) show expected fold changes. A recent study used calibrated editing detection to validate RNA editing sites, highlighting the importance of rigorous QC in transcriptomic analysis High-precision Detection of RNA Editing Sites using Calibrated Differential RNA Editing Scanner, J Vis Exp 2025.

Common Mistakes

Insufficient biological replication. Two replicates per condition cannot provide reliable p values and inflate false positives. Use at least three, and preferably more, biological replicates.

Ignoring batch effects. Process all samples in a randomized order in the same batch. If that is impossible, include batch as a covariate in your statistical model. A systematic multiomic study in donkey tissues found that batch correction was necessary to integrate different data types and avoid confounding Multiomic data integration unveils molecular mechanisms underlying chestnut and skin tissue differentiation in Dezhou donkeys, BMC Genomics 2025.

Over trimming reads. Trimming removes bases with low quality scores, but setting a very stringent threshold (e.g., quality < 30) can cut into unique sequence. Keep a minimum read length of 30 bp after trimming.

Mis specifying strandedness. If you do not know or incorrectly set the strandedness parameter during alignment or counting, you will lose most of your reads. Use a stranded specific library kit and note its orientation (usually RF or dUTP based).

Using the wrong reference annotation. Old or incomplete GTF files cause many reads to be assigned as intergenic. Always download the current annotation from Ensembl or GENCODE.

Relying solely on p values without effect size. Genes with very low counts and huge fold changes may be noise. Filter low expressed genes (e.g., requiring at least 10 counts in a minimum number of samples) and inspect the magnitude of change.

Limits or Uncertainty

RNA seq can quantify the transcriptome, but it has important limitations.

It measures relative, not absolute, abundance. Without spike in controls, you cannot compare expression levels across genes or across samples for fold changes beyond the normalization scale. The method assumes the total RNA population is similar between compared samples, which may not be true for diverse tissues or treatments.

It does not capture post transcriptional regulation directly. Changes in RNA abundance do not always correlate with protein levels due to translation efficiency, protein stability, and modifications. RNA seq also misses non polyadenylated RNAs unless rRNA depletion is used.

Splice variant quantification is limited by read length. Short reads (50 75 bp) can only resolve isoforms if they include unique splice junctions. Long read technologies improve this, but they have higher error rates and lower throughput, as noted in a benchmark of nanopore RNA seq methods A systematic benchmark of bioinformatics methods for single-cell and spatial RNA-seq nanopore long reads data, NAR Genom Bioinform 2025.

Detection of very lowly expressed genes is noisy. Sequencing depth and library complexity set the lower detectability threshold. Many genes with fewer than 5 counts in a sample are filtered out, potentially missing biologically relevant transcripts.

Genetic variation can confound mapping. Single nucleotide variants in the sample relative to the reference can cause misalignment, especially in non model organisms. Personalizing the genome for the sample improves accuracy.

Batch effects and technical confounders can dominate biological signal. Even with careful randomization, latent variables like RNA extraction batch or sequencing lane can introduce systematic bias. Computational correction (e.g., limma removeBatchEffect, sva) helps but cannot eliminate unknown confounders.

RNA seq cannot directly detect RNA editing events without specialized pipelines. Standard alignment ignores base mismatches that may represent edits. Dedicated tools are required to separate editing from sequencing errors and SNPs, as demonstrated in a protocol for calibrated editing detection High-precision Detection of RNA Editing Sites using Calibrated Differential RNA Editing Scanner, J Vis Exp 2025.

Frequently Asked Questions

Q: How many reads do I need for differential expression?
A: For gene level analysis in human or mouse, 10 to 30 million read pairs per sample is typical. The exact number depends on the expression distribution and the desired fold change detection. Use the Scotty tool or a power analysis from your pilot data.

Q: Should I remove ribosomal RNA or select polyA?
A: For mRNA focused studies on eukaryotes, polyA selection is simpler and cheaper. For total RNA studies or prokaryotes, use rRNA depletion. Recent methods like Ribo Zero or commercial kits work well.

Q: Can I combine two different library preparation protocols in one analysis?
A: It is not recommended because each protocol introduces different biases (e.g., 3 prime bias for polyA, GC bias for total RNA). If you must combine, use a normalization method that accounts for library type and include the protocol as a batch variable.

Q: How do I handle reads that map to multiple locations?
A: Multi mapping reads are common in repetitive regions and gene families. Most quantification tools assign them probabilistically (e.g., Salmon or RSEM). For standard differential expression, the default handling is acceptable. For gene families with high homology, consider using unique mappable reads only or family level collapsing.

References and Further Reading

• NCBI Bookshelf. Comprehensive overview of RNA seq technology and laboratory methods. NCBI Bookshelf
• EMBL EBI Training. RNA seq analysis training course covering quality control to differential expression. EMBL EBI Training
• Galaxy Training Network. Step by step tutorials for RNA seq analysis using open source tools. Galaxy Training Network
• Bioconductor. Statistical software packages for RNA seq normalization and differential expression. Bioconductor
• NCBI Sequence Read Archive. Public repository for raw sequencing data and metadata. NCBI Sequence Read Archive
• A systematic benchmark of bioinformatics methods for single-cell and spatial RNA-seq nanopore long reads data. NAR Genom Bioinform 2025. PubMed
• Dual RNA-seq Identifies Virulence Factors in Mesanophrys sp. and Immune Defenses in Portunus trituberculatus Hemocytes. Acta Parasitol 2025. PubMed
• High-precision Detection of RNA Editing Sites using Calibrated Differential RNA Editing Scanner. J Vis Exp 2025. PubMed
• Multiomic data integration unveils molecular mechanisms underlying chestnut and skin tissue differentiation in Dezhou donkeys. BMC Genomics 2025. PubMed
• Clinical Manifestations and Genetic Findings in Three Patients with Chediak-Higashi Syndrome: Highlighting the Splice Site Variants. Iran J Allergy Asthma Immunol 2025. PubMed
• AAV vector production in suspension cells using PEI transfection and sodium butyrate with orthogonal assessment of function and quality. Mol Ther Adv 2025. PubMed

Related Articles

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