RNA sequencing (RNA-seq) is a laboratory method that uses high-throughput sequencing to reveal the presence and quantity of RNA in a biological sample at a given moment. You start with total RNA, convert it into a library of complementary DNA (cDNA) fragments, and then sequence those fragments to produce millions of short reads that are mapped to a reference genome or transcriptome to measure gene expression levels.
Formula
RPKM (reads per kilobase of transcript per million mapped reads) is a normalized measure of gene expression from RNA-seq data. The formula is:
RPKM = (reads mapped to gene) / (gene length in kb × total mapped reads in millions)
Where:
- reads mapped to gene (r) – the number of sequencing reads that align uniquely to the gene of interest (count)
- gene length (g) – the length of the gene’s transcript in kilobases (kb), where 1 kb = 1,000 bases
- total mapped reads (m) – the total number of sequencing reads that successfully align to the reference genome or transcriptome, expressed in millions (e.g., 30 = 30,000,000 reads)
The output RPKM gives a normalized expression value that accounts for both sequencing depth (total reads) and gene length, allowing comparisons between genes within a sample.
Worked Example
Using the default values from the calculator:
- Total mapped reads = 30 million
- Gene length = 1.5 kb
- Reads mapped to gene = 1,200
Step 1: Multiply gene length by total mapped reads (in millions):
1.5 kb × 30 million = 45 kb·million
Step 2: Divide reads mapped to gene by that product:
1,200 / 45 = 26.6667
So RPKM = 26.67 (rounded to 4 decimal places as per the calculator output).
The RPM (reads per million) is simply 1,200 / 30 = 40.00.
Assumptions
- All reads counted are uniquely mapped to the gene (no multi-mapping reads).
- Gene length is the sum of all exons of the transcript, not including introns.
- Total mapped reads excludes reads that map to ribosomal RNA, mitochondrial RNA, or unmapped reads.
- The calculation assumes a single-end sequencing protocol; for paired-end reads, each fragment is counted once.
What RNA Sequencing Actually Measures
RNA-seq is not a single step but a multi-stage workflow. It begins with RNA extraction from cells or tissues, followed by removal of abundant ribosomal RNA (rRNA) or enrichment of messenger RNA (mRNA) using poly-A selection. The RNA is then fragmented into short pieces (typically 200–500 nucleotides), and reverse transcriptase converts these RNA fragments into cDNA. Adapters—short known DNA sequences—are ligated to both ends of the cDNA fragments, and the library is amplified by PCR. Finally, the library is loaded onto a sequencing flow cell, where millions of clusters are sequenced by synthesis (e.g., Illumina platforms) or by other chemistries, producing short reads (usually 50–150 bases).
From Reads to Expression Values
After sequencing, the raw reads are processed through a bioinformatics pipeline:
- Quality control – remove low-quality bases and adapter sequences using tools like FastQC or Trimmomatic.
- Alignment – map reads to a reference genome (e.g., human genome GRCh38) using splice-aware aligners like STAR or HISAT2, which can handle reads spanning exon-exon junctions.
- Quantification – count how many reads overlap each gene (using featureCounts or HTSeq).
- Normalization – apply methods like RPKM/FPKM (for gene length and depth) or TPM (transcripts per million), which is now preferred because it is more consistent across samples.
- Differential expression – compare counts between conditions using statistical models (e.g., DESeq2 or edgeR) to identify genes that change significantly.
The calculator above computes RPKM, a classic normalization that remains useful for understanding the concept, though modern pipelines often use TPM or raw counts with statistical modeling.
Common Problems and Misconceptions
A frequent mistake is confusing RPKM with TPM. RPKM normalizes by total reads and gene length, but the sum of RPKMs across all genes is not constant between samples, making cross-sample comparisons tricky. TPM normalizes by transcript length first, then scales to a per-million value, so the sum of TPMs is the same across samples. Another issue is ignoring multi-mapping reads—reads that align to multiple locations (e.g., paralogous genes) should be handled carefully, often by discarding them or assigning them fractionally. Also, ribosomal RNA contamination can dominate sequencing if not removed, wasting reads. Finally, low-input RNA samples (e.g., from single cells) require amplification, which introduces bias; unique molecular identifiers (UMIs) help correct for PCR duplicates.
Choosing the Right RNA-seq Method
Your choice depends on the biological question:
- mRNA-seq – poly-A selection, best for studying protein-coding genes.
- Total RNA-seq – removes rRNA but retains non-coding RNAs; better for lncRNAs and circRNAs.
- Small RNA-seq – size-selection for microRNAs and other small RNAs.
- Single-cell RNA-seq – isolates individual cells, often using droplet-based platforms (e.g., 10x Genomics), to capture cellular heterogeneity.
- Strand-specific RNA-seq – preserves the orientation of transcripts, which helps in detecting antisense transcription and accurately quantifying overlapping genes.
Each method has trade-offs in cost, input RNA amount, and bioinformatics complexity.
Costs and Time Considerations
RNA-seq is more expensive than qPCR but cheaper than whole-genome sequencing. Costs vary widely by depth (number of reads per sample), read length, and whether you use a core facility or commercial service. A typical mRNA-seq run for 10–20 million reads per sample might cost $100–$300 per sample in reagents, plus library prep kits and sequencing machine amortization. The time from RNA to data is roughly 2–5 days for library prep and sequencing, followed by a few hours to a day of bioinformatics analysis for a small dataset. Deeper sequencing (50–100 million reads) is needed for detecting low-abundance transcripts or isoforms.
Safety and Best Practices
RNA is highly susceptible to degradation by RNases, which are ubiquitous on skin and surfaces. Always wear gloves, use RNase-free tubes and water, and keep samples on ice. When working with chemicals like TRIzol (phenol-based), follow your lab method, the safety data sheet (SDS) and your supervisor. For bioinformatics, ensure your reference genome version matches your alignment tool’s requirements to avoid mapping errors. Follow your lab method, the safety data sheet (SDS) and your supervisor.
Frequently Asked Questions
What is the difference between RNA-seq and microarray? Microarrays use predefined probes to measure known transcripts, limiting discovery. RNA-seq sequences all RNA molecules, allowing detection of novel transcripts, splice variants, and mutations without prior knowledge.
How many reads do I need for RNA-seq? For differential expression of protein-coding genes, 10–30 million reads per sample is typical. For detecting rare transcripts or isoforms, you may need 50–100 million. Single-cell RNA-seq often uses fewer reads per cell (e.g., 10,000–50,000) because of the large number of cells.
What does RPKM stand for and why is it used? RPKM stands for Reads Per Kilobase of transcript per Million mapped reads. It normalizes for sequencing depth and gene length, allowing comparisons of expression levels between genes in the same sample. However, it is being replaced by TPM in many studies.
Can RNA-seq detect alternative splicing? Yes, if reads span exon-exon junctions and the sequencing depth is sufficient. Tools like rMATS or MISO can quantify isoform usage, but short reads (150 bp) may not fully resolve complex isoforms; long-read sequencing (e.g., Oxford Nanopore) can help.
Is RNA-seq quantitative? Yes, but it is relative—it measures the proportion of reads from each gene, not absolute copy numbers. Spike-in controls of known concentrations can be added to estimate absolute abundance.
Why do I get zero reads for a gene I know is expressed? Possible reasons: the gene is expressed at low levels below detection, reads failed to map due to mismatches, the gene is not annotated in your reference, or the RNA was degraded. Check your alignment statistics and quality control metrics.
Related
Formula used
in m | Total mapped reads (millions) | 30 | 1 | 1000 in g | Gene length (kb) | 1.5 | 0.1 | 100 in r | Reads mapped to gene | 1200 | 0 | 10000000 out rpkm | RPKM (reads per kb per million) | r / (g * m) | 4 out rpm | Reads per million | r / m | 2
Each line is one input (in, sel, list) or one result (out: label, unit, formula, decimals). How the calculators work
Method: the standard formula for this calculation with the constants and conversion factors written on the page, rounded as each result states. See method and constants, the disclaimer or report an error.
More in this section
- Calcium Nitrate Boiling Point
- Cation Formula for Magnesium Nitrate0.0788 mol/L
- Potassium Nitrate Solubility in Water: Full Data25.9 g/100 mL
- Greenhouse Grade Potassium Nitrate: Uses & Rates0.2000 mol/L
- Molar Mass of KNO2: Formula & Calculation0.1373 mol/L
- Potassium Nitride Formula & How to Write It