What is RNA-seq

What RNA-seq measures

RNA-seq measures transcript abundance across the entire transcriptome. Every gene that is actively being expressed in a cell produces messenger RNA. RNA-seq captures those mRNA molecules, converts them to DNA, and sequences them. The number of sequencing reads that map to a gene reflects how actively that gene is expressed.

This gives you a quantitative snapshot of gene expression at a specific moment, in a specific tissue, under specific conditions. You can compare these snapshots across conditions to find genes that are differentially expressed.

How short-read sequencing works

Most RNA-seq experiments today use short-read sequencing on Illumina platforms. The process has several steps.

RNA extraction. Total RNA is isolated from your biological sample. The quality of this RNA is critical. Degraded RNA will produce unreliable results. RNA integrity is measured using an Agilent Bioanalyzer or TapeStation, which produces an RNA Integrity Number. Aim for an RIN above 7.

Library preparation. The RNA is converted into a sequencing library. First, ribosomal RNA is removed or poly-A tailed mRNA is selected. Then the RNA is fragmented into short pieces, typically 200 to 500 nucleotides long. These fragments are reverse transcribed into complementary DNA. Short DNA sequences called adapters are ligated onto both ends of each fragment. These adapters serve two purposes: they allow the fragments to bind to the sequencing flow cell, and they provide priming sites for sequencing.

Cluster generation. The adapter-ligated fragments are loaded onto a glass flow cell. Each fragment binds to complementary oligonucleotides on the flow cell surface. Through a process called bridge amplification, each fragment is copied thousands of times to form a dense cluster of identical molecules. This amplification is necessary because a single molecule does not produce enough signal to detect.

Sequencing by synthesis. The sequencer reads each cluster one base at a time. In each cycle, fluorescently labeled nucleotides are added. Only the correct complementary nucleotide incorporates at each position. A camera captures the fluorescent signal, identifying which base was added. The fluorescent tag is then removed, and the next cycle begins. After hundreds of cycles, each cluster produces a read of known sequence.

This short video from Illumina shows the full process:

Single-end vs paired-end reads

In single-end sequencing, each fragment is read from one end only. You get one read per fragment.

In paired-end sequencing, each fragment is read from both ends. You get two reads per fragment, one from each direction. The two reads are called read 1 and read 2. The distance between them reflects the original fragment size.

Paired-end reads provide several advantages. They improve alignment accuracy because the aligner can use the expected distance between read pairs as a constraint. They help resolve ambiguous mappings in repetitive regions. They are essential for detecting structural variants and gene fusions.

For standard differential expression analysis, paired-end reads are the default choice. Single-end sequencing is cheaper and sometimes sufficient for simple gene-level quantification, but paired-end data is more versatile.

Read length choices

Illumina platforms produce reads of fixed length. Common choices are 50bp, 75bp, 100bp, and 150bp. The right choice depends on your application.

50bp reads are the shortest common option. They are sufficient for gene-level quantification when you have a well-annotated reference genome. They are the cheapest per read and produce the most reads per sequencing run. Many large-scale gene expression studies use 50bp reads.

75bp reads offer a good balance between cost and mapping quality. They are a common default for RNA-seq at many sequencing centers.

100bp reads improve mapping across splice junctions. They are helpful when you need to distinguish between transcript isoforms or when working with less well-annotated genomes.

150bp reads provide the best mapping quality. They are useful for novel transcript discovery, alternative splicing analysis, and de novo transcriptome assembly where no reference genome is available.

For a typical differential expression experiment with a well-annotated model organism, 75bp paired-end reads are a solid choice. Longer reads add cost without much benefit for simple gene-level counting.

Typical output: FASTQ files

The sequencer produces FASTQ files. Each read in a FASTQ file has four lines:

@SRR1234567.1 1/1
ATCGATCGATCGATCGATCGATCGATCGATCGATCGATCGATCGATCG
+
IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

• Line 1 is the read identifier, starting with @.
• Line 2 is the nucleotide sequence.
• Line 3 is a separator, always +.
• Line 4 is the quality string. Each character encodes the confidence that the corresponding base was called correctly.

A typical RNA-seq experiment produces 20 to 40 million reads per sample. With paired-end sequencing, each sample produces two FASTQ files: one for read 1 and one for read 2. A single compressed FASTQ file for one sample is usually 2 to 5 GB.

RNA-seq vs microarrays

Before RNA-seq, gene expression was measured using microarrays. A microarray is a glass slide with thousands of known DNA sequences attached at fixed positions. Fluorescently labeled cDNA from your sample hybridizes to these probes, and the fluorescence intensity indicates expression level.

Microarrays served the field well for over a decade, but they have fundamental limitations that RNA-seq overcomes.

Dynamic range. Microarray signals saturate at high expression levels and are noisy at low levels. RNA-seq has a much wider dynamic range. It can detect both very lowly expressed and very highly expressed genes accurately.

No prior knowledge required. Microarrays can only measure genes that have probes designed for them. RNA-seq sequences everything, so it can detect novel transcripts, unannotated genes, and non-coding RNAs without any prior knowledge.

No cross-hybridization. Microarray probes sometimes bind to similar but incorrect sequences. This produces false signals. RNA-seq reads are aligned computationally, and ambiguous alignments can be identified and handled.

Splice variant detection. Microarrays struggle to distinguish between different splice variants of the same gene. RNA-seq reads span exon junctions, making isoform-level analysis possible.

Quantitative accuracy. RNA-seq read counts are proportional to transcript abundance. This makes the data inherently more quantitative than fluorescence intensity values.

Today, RNA-seq has almost entirely replaced microarrays for gene expression profiling. The cost of sequencing has dropped to the point where RNA-seq is comparable in price to microarrays while providing far richer data.

Next steps

Now that you understand what RNA-seq measures and how the data is generated, the next page covers Experimental Design.