RNA sequencing

RNA-seq begins with the isolation of total RNA from a sample, followed by enrichment for messenger RNA (mRNA) using poly-A selection or depletion of ribosomal RNA. The RNA is then fragmented and reverse-transcribed into complementary DNA (cDNA), which is ligated to adapters and amplified. The resulting library is sequenced on a high-throughput platform, producing millions of short reads that are aligned to a reference genome or transcriptome, or assembled de novo. The number of reads mapping to each gene provides a digital measure of its expression level, typically normalized as fragments per kilobase of transcript per million mapped reads (FPKM) or transcripts per million (TPM).¹ This workflow can be adapted for single-cell RNA-seq, where individual cells are barcoded and pooled, enabling transcriptomic profiling at cellular resolution.²

RNA-seq has transformed the study of gene regulation, enabling the identification of differentially expressed genes across conditions, tissues, or developmental stages. It has been instrumental in cancer research, where it reveals fusion genes, somatic mutations in expressed genes, and aberrant splicing patterns that drive tumorigenesis.³ In clinical settings, RNA-seq is used for the molecular classification of tumors, guiding targeted therapies. Beyond mRNA, RNA-seq variants such as small RNA-seq and total RNA-seq capture microRNAs, long non-coding RNAs, and circular RNAs, expanding the known repertoire of regulatory molecules. The technique also underpins single-cell transcriptomics, which has uncovered cell-type heterogeneity in tissues and identified rare cell populations in the brain, immune system, and tumors.⁴

The analysis of RNA-seq data is computationally intensive, involving quality control, read alignment, quantification, and statistical testing. A major challenge is the accurate mapping of reads that span exon–exon junctions, which requires splice-aware aligners such as STAR or HISAT2. Quantification methods like Salmon and kallisto use lightweight algorithms to estimate transcript abundances without full alignment, improving speed and memory usage.⁵ Batch effects, library preparation biases, and low-input samples can introduce technical variability, necessitating robust experimental design and normalization methods such as TMM or DESeq2's median-of-ratios. Despite these hurdles, RNA-seq remains the gold standard for transcriptome analysis, with ongoing improvements in long-read sequencing (e.g., Oxford Nanopore) enabling full-length transcript isoform detection.⁶

Beyond standard gene expression, RNA-seq has revealed pervasive transcription: a large fraction of the genome is transcribed into non-coding RNAs, many with unknown functions. RNA editing events, such as A-to-I changes, can be detected by comparing RNA-seq reads to the genome, providing insights into post-transcriptional regulation.⁷ In metagenomics, RNA-seq (meta-transcriptomics) profiles the active gene expression of microbial communities without prior knowledge of their composition. A niche application is the use of RNA-seq to infer the evolutionary conservation of splicing patterns across species. Additionally, RNA-seq data can be re-analyzed to detect allele-specific expression, revealing genomic imprinting and cis-regulatory effects. The technique has also been applied to ancient RNA from archaeological specimens, though degradation limits its utility.⁸

Transcriptome

The complete set of RNA transcripts produced by the genome under specific circumstances.

cDNA

Complementary DNA synthesized from an RNA template by reverse transcriptase.

FPKM

Fragments per kilobase of transcript per million mapped reads; a normalization unit for RNA-seq.

TPM

Transcripts per million; a normalization method that scales for transcript length and library size.

Single-cell RNA-seq

A technique that profiles the transcriptome of individual cells, revealing cellular heterogeneity.

Splice-aware alignment

Read alignment that accounts for introns, allowing reads to span exon junctions.