What Bulk RNA-Seq Costs Per Sample in 2026
A bulk RNA-seq sample in 2026 costs between roughly $40 and $250 for library preparation plus sequencing, and the spread is almost entirely explained by read depth and library chemistry. At the low end you get a 3’ tag-based library with 5 to 10 million single-end reads, suitable for counting moderately expressed genes across many samples. At the high end you get a full-length poly(A) or rRNA-depleted library with 50 to 100 million paired-end reads, which supports isoform-level work and detection of low-abundance transcripts. Add RNA extraction, quality control, and analysis and the all-in figure per sample typically lands between $120 and $400.
The advertised numbers you see ranking on search ($109 promotional, “$36.9 to $173”) are real but incomplete. They price a library and a slice of a flow cell. They do not price extraction from your tissue type or failed samples. Also excluded are replicate counts large enough for the statistics to hold and the person who runs the analysis. This page separates those layers so you can build a budget that survives contact with the data.
What the per-sample price is made of
Four cost components sit behind any quote, and they scale differently.
RNA extraction and quality control runs $20 to $60 per sample at a service provider. Blood is the awkward case because globin messenger RNA can consume a large share of your reads, so providers add a globin depletion step for $15 to $30. Quality control means a fluorometric concentration measurement and a fragment analyzer trace giving an RNA integrity number (RIN), a 1-to-10 score of how degraded the RNA is. Below about RIN 7 a full-length poly(A) protocol starts producing 3’ coverage bias, and below RIN 5 you should move to a degraded-RNA chemistry instead.
Library preparation is the largest fixed cost and ranges from about $25 for a 3’ tag-based kit to $100 for rRNA depletion. Poly(A) selection captures polyadenylated messenger RNA and costs roughly $50 to $70 per sample. Ribosomal RNA depletion costs more, typically $80 to $110, and buys you non-polyadenylated species including most long non-coding RNAs and nascent transcripts. If your samples are FFPE tissue or otherwise degraded, rRNA depletion is not optional.
Sequencing is priced per gigabase and falls fast with instrument scale. On a NovaSeq X 25B flow cell, a service provider’s marginal cost per gigabase sits under $2, so 30 million 100 bp paired-end reads (about 6 Gb) can be sold profitably for $15 to $25. On a NextSeq 2000 P3 flow cell the per-gigabase cost is roughly four to six times higher, which is why small-batch quotes look expensive relative to core facility rates on the large instruments.
Analysis is the component most often left out. A standard pipeline is cheap in compute and expensive in attention. Salmon or kallisto quantification against a GENCODE transcriptome takes about 10 minutes and 4 GB of memory per sample on a laptop-class machine. A spot cloud instance handling 24 samples costs a few dollars. Interpreting the result is where the hours go.
Depth: how many reads you need
Read depth is the dial that moves price most, and the right setting depends entirely on the question.
For differential expression across protein-coding genes, 20 to 30 million reads per sample is the standard recommendation and the returns flatten quickly after that. Adding reads mostly improves detection of transcripts already near the noise floor. Single-end 50 to 100 bp reads are sufficient for gene-level counting, and paired-end reads buy you little at this scale beyond better handling of multi-mapping regions.
For isoform quantification and splicing, you need paired-end reads and 50 to 100 million of them. Distinguishing two isoforms that differ by one exon requires reads spanning the junction, and junction coverage is a small fraction of total coverage. This is the case where the original argument for RNA-seq over microarrays is sharpest, because sequencing resolves transcript structure rather than reporting a single probe intensity per gene.1
For novel transcript discovery or fusion detection, budget 100 million paired-end reads and expect to run a spliced aligner such as STAR rather than a transcriptome-only quantifier. That pushes you to the top of the price band.
Given a fixed budget, we would spend it on replicates rather than depth. Recent work on small cohorts shows that differential expression and enrichment results are poorly replicable when group sizes are small, with substantial disagreement in gene lists between independent samples drawn from the same population.2 Three versus three is a pilot. It is not a result. Six versus six is the minimum we would design for, and the increase is cheaper than most people expect because the sequencing component is the smallest line item.
Platform economics: why the same experiment has three prices
The instrument behind your quote explains most of the variance between vendors, so it is worth knowing what those machines cost.
A NovaSeq 6000 listed around $985,000 new, and the NovaSeq X Plus that replaced it sits near $1.25 million. Those instruments only make sense at high, steady throughput, which is why they live in core facilities and large commercial labs. Their output per run is enormous, so the per-sample sequencing cost is low but the batch size needed to fill a flow cell is large. Small orders get pooled with other customers’ samples, which is fine, but it means turnaround is set by when the queue fills.
A NextSeq 2000 costs roughly $335,000 and a NextSeq 1000 about $210,000. The per-gigabase cost is higher, the flow cells are smaller, and a single lab can fill one with its own project. If you need a 24-sample experiment back in ten days, this is the class of instrument that will do it.
Element Biosciences AVITI and MGI DNBSEQ instruments have pushed the mid-range price down further and now compete directly on per-gigabase cost. Their data is usable with the same tooling, since everything downstream reads FASTQ.
The practical consequence is simple. If your project is 8 samples, expect $200 to $350 per sample and pay for the turnaround. If your project is 96 samples, negotiate against the large-instrument rate and expect $80 to $140 all-in.
What to ask for in the deliverable
The quote should specify deliverables, and this is where you protect yourself. Ask for raw FASTQ files with the original per-base quality scores. A processed count matrix alone is not enough. Ask for the index sequences and the demultiplexing report so you can check for index hopping, which is a real artifact on patterned flow cells. Ask for the exact library kit name and the spike-in controls if any were used.
A count matrix alone is a weaker deliverable because you cannot re-quantify against a different annotation or check the failure modes. Technical variation in bulk transcriptomic data is substantial, and batch effects, library composition, and platform differences all leave signatures that only surface when you can inspect the raw reads.3 If you plan to combine your data with a public dataset, you will need to model those effects explicitly.
For the analysis itself, our default is Salmon with --validateMappings --seqBias --gcBias against a decoy-aware GENCODE index, then tximport into DESeq2. If you work in Python rather than R, InMoose ports the core limma and edgeR differential expression and batch correction methods into Python with results that match the R implementations closely.4 That removes the main reason people used to keep a separate R environment just for transcriptomics.
Where RNA-seq sits relative to other assays
Cost comparisons only make sense against what the assay tells you. A transcriptome is a snapshot of regulatory state at the moment of collection, which makes it more informative about current biology than a genome and far more variable between draws.5 Time of day, fasting state, recent infection, and exercise all move gene expression.
That variability is the argument for treating RNA-seq as a repeated measurement rather than a one-time test. It is also the argument for pairing it with assays on different timescales. Integrating transcriptomics with proteomics and chromatin data resolves regulatory relationships that neither layer shows alone, because messenger RNA abundance and protein abundance often disagree.6 Methods for combining these layers are now standard practice rather than a research novelty.7
Interpreting any of this in the context of your own health requires a clinician. We describe measurement and what the numbers mean statistically, and nothing here is a diagnosis.
Questions people also ask
How much does RNA sequencing typically cost? Library preparation plus sequencing runs $40 to $250 per sample depending on depth and chemistry. Including extraction, quality control, and analysis, budget $120 to $400 per sample for a small project and $80 to $140 for a project of 96 samples or more.
How much does a NovaSeq 6000 cost? The list price was approximately $985,000 when it was Illumina’s flagship, and its successor the NovaSeq X Plus is priced near $1.25 million. Reagent cost per run is the larger long-term expense, and filling the flow cell is what makes the economics work.
How much does a NextSeq 2000 cost? Roughly $335,000 for the instrument, with the NextSeq 1000 near $210,000. Per-gigabase reagent cost is several times the NovaSeq X rate, which is the price you pay for smaller flow cells and faster turnaround.
How much will whole genome sequencing cost in 2026? Commercial 30x human whole-genome sequencing sits in the $200 to $600 range for the sequencing itself, with clinical-grade interpretation costing considerably more. The steady fall in cost per base has been the driving force behind every application of short-read sequencing since 2007.8
Is it cheaper to buy your own sequencer? Almost never for RNA-seq. A benchtop instrument only breaks even at sustained high utilization, and idle reagent expiry plus staff time usually exceeds the premium a service provider charges.
What annotation should I quantify against? GENCODE for human, matched to a specific release number that you record. Switching annotation versions mid-project changes gene-level counts enough to shift borderline differential expression calls.
Footnotes
-
Zhong Wang, Mark Gerstein, Michael Snyder. RNA-Seq: a revolutionary tool for transcriptomics. Nature Reviews Genetics, 2009. https://doi.org/10.1038/nrg2484 ↩
-
Peter Methys Degen, Matúš Medo. Replicability of bulk RNA-Seq differential expression and enrichment analysis results for small cohort sizes. PLOS Computational Biology, 2025. https://doi.org/10.1371/journal.pcbi.1011630 ↩
-
Hengrui Liu, Yiying Li, Miray Karsidag, et al. Technical and Biological Biases in Bulk Transcriptomic Data Mining for Cancer Research. Journal of Cancer, 2025. https://doi.org/10.7150/jca.100922 ↩
-
Maximilien Colange, Guillaume Appé, Léa Meunier, et al. Bridging the gap between R and Python in bulk transcriptomic data analysis with InMoose. Scientific Reports, 2025. https://doi.org/10.1038/s41598-025-03376-y ↩
-
Erica V. Todd, Michael A. Black, Neil J. Gemmell. The power and promise of RNA ‐seq in ecology and evolution. Molecular Ecology, 2016. https://doi.org/10.1111/mec.13526 ↩
-
Marco Cavalli, Klev Diamanti, Gang Pan, et al. A Multi-Omics Approach to Liver Diseases: Integration of Single Nuclei Transcriptomics with Proteomics and HiCap Bulk Data in Human Liver. OMICS: A Journal of Integrative Biology, 2020. https://doi.org/10.1089/omi.2019.0215 ↩
-
Biswapriya B Misra, Carl Langefeld, Michael Olivier, et al. Integrated omics: tools, advances and future approaches. Journal of Molecular Endocrinology, 2019. https://doi.org/10.1530/jme-18-0055 ↩
-
Jerzy K. Kulski. Next-Generation Sequencing — An Overview of the History, Tools, and “Omic” Applications. Next Generation Sequencing - Advances, Applications and Challenges, 2016. https://doi.org/10.5772/61964 ↩
Pothos builds longitudinal molecular profiles of individuals: whole-genome sequencing, single-cell RNA sequencing, and plasma proteomics, delivered as a detailed report, every data file, and stored cells. Build your profile.