What Is RNA Sequencing and Transcriptome Analysis?
RNA sequencing, commonly called RNA-seq, is a high-throughput approach for investigating RNA molecules present in a biological sample.
The transcriptome comprises the RNA transcripts expressed in a cell population, tissue or other sample under particular conditions. RNA-seq can support the measurement of gene expression, transcript abundance and selected RNA-processing events.
In a common workflow, RNA is extracted, prepared into a sequencing library and sequenced. Computational methods then relate the resulting reads to genomic or transcript reference sequences.
What questions can RNA-seq answer?
- Which genes are expressed in a sample?
- Which genes differ in expression between conditions?
- How do transcriptional profiles vary across tissues?
- Are particular pathways associated with observed expression changes?
- Which transcript isoforms are supported by the data?
RNA abundance is a measurement of transcripts, not a direct measurement of protein concentration, enzymatic activity or cellular function.
RNA-Seq Experimental Design and Biological Replicates
Reliable differential expression analysis begins before sequencing. The study design determines which biological comparisons can be estimated and how uncertainty should be modeled.
Biological vs. technical replicates
Biological replicates represent distinct biological units, such as independently collected samples. They provide information about biological variability.
Technical replicates repeat a measurement or processing step on the same biological material. They can help characterize technical variation but do not replace independent biological replication.
| Design factor | Why it matters |
|---|---|
| Biological replication | Supports estimation of variation between biological units |
| Batch structure | Can introduce technical differences unrelated to the biological question |
| Sample randomization | Helps reduce confounding between study groups and processing conditions |
| RNA integrity | May influence library composition and quantitative results |
| Sequencing depth | Affects information available for low-abundance transcripts and other tasks |
| Library protocol | Determines which RNA species and transcript regions are represented |
A design that completely confounds treatment with sequencing batch may make their effects impossible to distinguish statistically. Data analysis cannot always repair such problems after sequencing.
FASTQ Quality Control in RNA-Seq
RNA-seq computational analysis generally begins with quality assessment of the raw sequencing reads and the associated sample metadata.
Common quality checks
- Per-base sequencing quality and read length.
- Adapter content and overrepresented sequences.
- Sequence duplication and library complexity.
- Potential contamination.
- GC-content distributions.
- Consistency of sample identifiers and library metadata.
RNA-seq libraries can have distinctive sequence composition because transcripts differ greatly in abundance. Some quality-control warnings may reflect the expected assay biology rather than a technical failure.
Post-alignment RNA-seq checks
After alignment or quantification, researchers may assess mapping rates, strandedness consistency, ribosomal RNA content, gene body coverage and other assay-relevant indicators.
Different RNA-seq protocols have different expectations, so quality thresholds should not be treated as universally applicable.
Evaluate quality across the complete workflow, not only at the raw FASTQ stage. An apparently high-quality library can still have sample-label, reference or biological interpretation problems.
Splice-Aware Alignment and Transcript Quantification
RNA-derived reads frequently span exon-exon junctions because many mature transcripts have undergone RNA splicing. This makes RNA-seq mapping different from ordinary contiguous DNA alignment.
Genome-based alignment
Splice-aware aligners such as STAR can map reads across exon junctions, supporting gene expression analysis and investigation of transcript structure.
Transcriptome-based quantification
Tools such as Salmon estimate transcript abundance using computational strategies that can avoid the need for conventional full genomic alignment in many workflows.
| Approach | Example | Output |
|---|---|---|
| Splice-aware genome alignment | STAR | Genomic read alignments |
| Transcript abundance estimation | Salmon | Transcript-level abundance estimates |
| Gene-level counting | featureCounts | Gene-by-sample count matrix |
RNA-Seq Read Counts and Gene Expression Matrices
One common output of an RNA-seq workflow is a count matrix. Rows represent genes and columns represent biological samples. Each cell contains a count or an estimated count associated with a particular gene.
Illustrative count matrix
| Gene | Control 1 | Control 2 | Treatment 1 | Treatment 2 |
|---|---|---|---|---|
| GENE_A | 120 | 132 | 302 | 289 |
| GENE_B | 845 | 811 | 402 | 427 |
| GENE_C | 54 | 61 | 57 | 65 |
| GENE_D | 240 | 228 | 245 | 231 |
These are fictional values for illustration. They should not be interpreted as real experimental findings.
Gene counts vs. transcript abundance
Gene-level counts aggregate evidence at the gene level. Transcript-level estimates attempt to distinguish transcript isoforms, which can be challenging when several isoforms share many exons.
Counts, TPM values and other abundance measurements are not interchangeable. Statistical methods designed for raw or appropriately estimated count inputs should receive the data type specified by their documentation.
Why RNA-Seq Normalization Is Essential
Raw counts are influenced by factors beyond biological expression, including sequencing depth and RNA composition. Normalization helps account for selected technical differences among samples.
Common normalization concepts
| Method or value | Primary purpose | Key consideration |
|---|---|---|
| Raw counts | Count-based statistical modeling | Not directly comparable without appropriate modeling |
| DESeq2 size factors | Account for between-sample scaling differences | Based on a median-of-ratios approach |
| TMM normalization | Adjust sample composition effects | Used in edgeR workflows |
| TPM | Length- and library-scaled expression summary | Not a substitute for raw counts in every statistical model |
| Variance-stabilizing transformation | Support exploration and visualization | Distinct from count-model input preparation |
Different normalization methods make different assumptions. Strong global expression shifts, compositional effects and study design can influence their performance.
Do not automatically replace gene counts with TPM values before running DESeq2. Use the count-based inputs and normalization strategy appropriate to the selected statistical framework.
DESeq2, edgeR and Statistical Gene Expression Analysis
Differential expression analysis tests whether RNA abundance differs systematically between specified biological conditions.
DESeq2 and edgeR use count-based statistical approaches designed to account for biological variability and the dispersion often observed in RNA-seq data.
Key differential expression outputs
| Output | Interpretation |
|---|---|
| log2 fold change | Estimated direction and magnitude of expression difference on a logarithmic scale |
| Standard error | Estimated uncertainty associated with an effect estimate |
| p-value | Evidence against a specified statistical null hypothesis |
| Adjusted p-value | Multiple-testing-adjusted significance measure |
| Base mean | Mean normalized expression level across analyzed samples in DESeq2 output |
What does log2 fold change mean?
A log2 fold change of +1 corresponds to a twofold increase in the modeled expression ratio. A value of −1 corresponds to a twofold decrease.
Statistical significance depends on effect size, sample variability, replication, modeling assumptions and multiple-testing adjustment. A significant gene is not automatically biologically important.
PCA Plots, Volcano Plots, Heatmaps and MA Plots
Visualization helps researchers examine sample relationships, detect unexpected patterns and communicate differential expression results.
Principal component analysis
Summarizes major patterns of variation among samples after appropriate transformation.
Volcano plot
Displays an effect-size estimate against a measure of statistical significance.
Expression heatmap
Represents selected gene expression values using a color scale.
MA plot
Displays expression differences as a function of mean abundance.
What does a volcano plot show?
A typical volcano plot displays log2 fold change on the horizontal axis and negative log10 p-value or adjusted p-value on the vertical axis. Points far from zero may indicate large estimated expression effects, while higher points indicate stronger statistical evidence under the chosen measure.
PCA separation may reflect treatment, batch, tissue type or other factors. A heatmap can look dramatically different depending on gene selection, scaling and clustering choices.
Visualization should therefore complement a prespecified statistical analysis rather than replace it.
Gene Ontology, Pathway Analysis and Functional Enrichment
Differentially expressed genes are often interpreted using biological annotations, gene sets and pathway databases. These resources can help researchers identify processes associated with observed expression patterns.
Common analytical approaches
| Approach | Purpose | Important consideration |
|---|---|---|
| Overrepresentation analysis | Tests whether selected genes are enriched for an annotation | Depends on gene-selection rules and background universe |
| Rank-based enrichment | Uses ordered gene-level statistics | Ranking metric and gene-set composition matter |
| Pathway analysis | Places findings in curated biological pathways | Annotation completeness and directionality vary |
Useful biological resources
- Gene Ontology — structured biological process, molecular function and cellular component annotations.
- Reactome — curated biological pathways.
- MSigDB — collections of annotated gene sets.
Functional enrichment suggests associations with annotated processes. It does not prove that a pathway is mechanistically activated or inhibited. Results may be affected by annotation coverage, gene dependencies and the selection of the statistical universe.
A coherent scientific narrative should connect differential expression results to study design, known biology, pathway evidence and proposed validation experiments.
RNA-Seq Interpretation, Reproducibility and Frequently Asked Questions
RNA-seq results depend on sample quality, library preparation, reference resources, quantification methods, normalization, statistical design and biological context.
A reproducible analysis should record sample metadata, reference versions, software parameters, count-generation procedures, statistical contrasts and decisions affecting reported genes.
Frequently asked questions
What is the main purpose of RNA-seq analysis?
To characterize RNA abundance and examine transcriptional differences among biological samples or experimental conditions.
What is the difference between RNA-seq and DNA-seq?
DNA-seq investigates DNA sequence information. RNA-seq investigates RNA molecules and their abundance or processing.
What does DESeq2 do?
DESeq2 provides statistical methods for differential analysis of count data, including dispersion and fold-change estimation.
Why is RNA-seq normalization needed?
Normalization accounts for selected between-sample technical differences, helping support meaningful statistical comparisons.
What is a volcano plot in RNA-seq?
A volcano plot displays estimated expression changes against a measure of statistical significance.
Does significant differential expression prove a gene causes a phenotype?
No. Differential expression identifies statistical associations in transcript abundance. Causal interpretation requires additional evidence and often experimental validation.
Can RNA-seq identify alternative splicing?
Yes, with suitable library designs, sequencing characteristics and specialized computational methods. Gene-level differential expression alone does not comprehensively characterize alternative splicing.
Scientific references
- Love M.I., Huber W. and Anders S. (2014). Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology. DOI: 10.1186/s13059-014-0550-8 .
- Robinson M.D., McCarthy D.J. and Smyth G.K. (2010). edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. DOI.
- Dobin A. et al. (2013). STAR: ultrafast universal RNA-seq aligner. Bioinformatics. DOI.
- Patro R. et al. (2017). Salmon provides fast and bias-aware quantification of transcript expression. Nature Methods. DOI.
- RNA-seq Data: Challenges in and Recommendations for Experimental Design and Analysis .
- From RNA-seq Reads to Differential Expression Results .
RNA-seq analysis transforms sequencing reads into quantitative evidence about transcription. Sound experimental design, suitable normalization, statistical uncertainty assessment and reproducible workflows are essential for reliable biological interpretation.