RNA-seq (short for RNA sequencing) is a method for reading the complete set of RNA molecules in a sample of cells or tissue. It reveals which genes are active, how active they are, and which versions of each gene are being produced. Because RNA changes rapidly in response to time of day, diet, disease, and countless other factors, RNA-seq essentially captures a snapshot of what your cells are doing at a specific moment.
How RNA-Seq Works
Every cell in your body contains the same DNA, but not every gene is switched on at the same time. When a gene activates, the cell copies that stretch of DNA into RNA, a temporary molecular messenger that carries instructions for building proteins. The full collection of RNA in a cell at any given moment is called the transcriptome, and RNA-seq is the most widely used way to measure it.
The process follows a consistent series of steps. First, researchers extract all the RNA from their sample, whether that’s a blood draw, a tumor biopsy, or a batch of lab-grown cells. Next, because sequencing machines read DNA rather than RNA, the extracted RNA is copied back into DNA (called cDNA). These cDNA molecules are then broken into small fragments, and short adapter sequences are attached to the ends of each fragment so the sequencing machine can recognize and read them. Finally, the prepared fragments are loaded onto a sequencer, which reads millions of them simultaneously.
The raw output is millions of short “reads,” each one a snippet of sequence from one of the original RNA molecules. Software maps these reads back to a reference genome to figure out which gene each snippet came from and tallies how many reads landed on each gene. More reads on a gene means that gene was more active in the sample.
What RNA-Seq Can Tell You
The most common use is differential gene expression analysis: comparing two groups of samples (say, healthy tissue versus cancerous tissue) to identify which genes are significantly more or less active. This is the backbone of thousands of studies in cancer biology, immunology, neuroscience, and drug development.
But RNA-seq goes well beyond simple on/off measurements. A single gene can produce multiple versions of its RNA by including or skipping certain segments, a process called alternative splicing. RNA-seq can detect these different versions, or isoforms, and determine whether one version becomes more common in disease. Researchers can also use the data to discover entirely new transcripts that weren’t in existing gene databases, identify genes producing very low levels of RNA, and study non-coding RNAs that regulate other genes without building proteins.
Why It Replaced Microarrays
Before RNA-seq, the standard tool for measuring gene activity was the microarray, a chip covered with thousands of pre-designed probes that bind to known gene sequences. Microarrays worked, but they had real limitations. They could only detect genes someone had already identified and designed probes for, they struggled with background noise from cross-binding, and they couldn’t reliably distinguish between “no expression” and “low expression.”
A head-to-head comparison in activated immune cells showed the difference clearly. RNA-seq demonstrated a dynamic range roughly 70 times broader than microarrays (260,000-fold versus 3,600-fold), meaning it could accurately measure both the quietest and loudest genes in the same experiment. In one case, RNA-seq detected a gene expressed at very low levels whose activity dropped by 97% after cell activation. The microarray missed the change entirely. In another, a commonly used reference gene appeared stable on the microarray but actually increased two to four-fold when measured by RNA-seq. That broader sensitivity is why RNA-seq has become the default for transcriptome profiling.
Bulk Versus Single-Cell RNA-Seq
Standard (bulk) RNA-seq grinds up a tissue sample and measures the average gene activity across all the cells in it. This works well when the sample is relatively uniform, but tissues are usually a mix of many cell types. A tumor biopsy might contain cancer cells, immune cells, blood vessel cells, and connective tissue, and their gene activity gets blended into one average signal.
Single-cell RNA-seq (scRNA-seq) solves this by isolating individual cells and sequencing each one separately. This reveals which cell types are present, how they differ from each other, and even identifies rare populations that would be invisible in a bulk measurement. It has become essential for mapping cell atlases, understanding how tumors evade the immune system, and tracing how stem cells develop into specialized tissues. The tradeoff is cost and complexity: single-cell experiments generate far more data and require specialized computational tools to analyze.
Short Reads Versus Long Reads
Most RNA-seq today uses short-read sequencing, which reads fragments roughly 150 base pairs long. This is fast, accurate, and relatively affordable, but it creates a puzzle: you’re trying to reconstruct full-length RNA molecules (often thousands of base pairs) from tiny overlapping pieces. That reconstruction step can fail for complex genes that produce many similar isoforms, because the short fragments don’t span enough of the molecule to tell the versions apart.
Long-read sequencing platforms can read entire RNA molecules from end to end, sometimes tens of thousands of base pairs in a single pass. This makes it far easier to identify exactly which combination of gene segments a particular isoform contains, to annotate genes more accurately, and to discover new types of non-coding RNA. Long-read accuracy has improved dramatically in recent years, though short-read sequencing still dominates for straightforward gene expression comparisons where its lower cost and higher throughput are more practical.
How Sequencing Depth Matters
Sequencing depth refers to how many reads you generate per sample. More reads means more chances to detect rare transcripts, but also higher costs. For a typical differential expression experiment, 50 to 150 million reads per sample is standard. This is enough to reliably quantify most genes that are active at moderate to high levels.
For specialized applications, depth requirements change significantly. Researchers at Baylor College of Medicine found that pushing to ultra-high depth (up to one billion reads per sample) uncovered low-abundance transcripts and rare splicing events that standard depths missed entirely. This proved particularly valuable for diagnosing genetic disorders where the disease-causing defect shows up as a subtle change in RNA processing. On the cost side, a standard mRNA sequencing library preparation currently runs around $325 per sample at academic core facilities, with sequencing costs on top of that.
Known Biases to Understand
RNA-seq is powerful, but it’s not a perfect mirror of biology. One well-documented bias involves transcript length. Because the process fragments RNA before sequencing, longer genes naturally produce more reads than shorter genes of the same activity level. This isn’t a flaw in any particular machine or protocol; it’s a mathematical consequence of cutting longer molecules into more pieces. The practical result is that statistical tests have more power to detect changes in long genes than short ones, which can skew results when comparing groups of genes with different average lengths.
Other sources of noise include variation in how efficiently different RNA sequences get copied into cDNA (partly influenced by the proportion of G and C bases in the sequence) and batch effects, where samples processed on different days or by different technicians show systematic differences unrelated to biology. Computational methods exist to correct for all of these, but they require careful experimental design. Running all comparison samples in the same batch and including enough biological replicates are two of the simplest ways to keep results reliable.
The Analysis Pipeline
After sequencing, the data passes through a series of computational steps. Quality control comes first: software scans the raw reads for errors, adapter contamination, and low-quality base calls, then trims or discards problematic reads. The cleaned reads are then aligned to a reference genome or transcriptome, which determines where each read originated. Finally, the software counts how many reads map to each gene, producing a table of raw counts that forms the basis for all downstream statistics.
From there, the analysis branches depending on the question. Differential expression analysis normalizes the counts to account for differences in sequencing depth between samples, then uses statistical models to identify genes with significantly different activity between conditions. Splicing analysis looks at how reads distribute across different segments of a gene to detect changes in isoform usage. Transcript discovery tools assemble reads into full-length structures and compare them against known gene models to flag new or unannotated forms. Each of these paths has its own specialized software and statistical considerations, which is why RNA-seq projects typically involve both wet-lab biologists and bioinformaticians working together.