Sequencing depth can dominate a cheap platform
A low library-prep price can buy little if reads land on empty droplets, doublets, ambient RNA, or insufficiently resolved cells.
What does a captured, sequenced cell actually cost?
The model separates library prep and reagent cost per cell from sequencing cost, which scales with how deeply each cell is read. A cheap capture platform can still produce an expensive experiment at high sequencing depth.
$1,340 total across 67,000 cells
- Library prep & reagents
- $0.010/cell
- Sequencing
- $0.010/cell
- Reads per cell
- 10,000
A 2026 HyDrop v2 study reported €668 (≈$700) for a 67,000-cell atlas excluding sequencing, about €0.01 per cell, roughly 14× below its quoted 10x v2 comparator. Illumina NovaSeq X flow-cell pricing runs approximately $1–1.6 per million reads depending on flow-cell size, cited here as an approximate $/€ 1:1 illustration, not an exchange-rate claim.
Calculation and boundaries
Cost per cell = library-prep/reagent cost per cell + (reads per cell ÷ 1,000,000) × sequencing price per million reads. Total cost = cost per cell × cells captured. Excludes tissue acquisition, labor, instrument amortization, compute and storage, and downstream analysis, all separate layers in the article's measurement stack. This is not the HyDrop v2 study's own cost breakdown, which reports one aggregate reagent figure rather than separating sequencing by depth.
Four stages, each capable of discarding information
A dataset is only as good as its weakest stage: deep sequencing cannot repair a biased specimen, and a perfect specimen is wasted by shallow sequencing.
Specimen
Collection, preservation, dissociation, nuclei isolation, and viability define what enters the assay.
- Measure
- Representative viable cells
- Failure boundary
- Selection and handling bias determine which cell states even reach the instrument.
Where the frontier moves
Preservation methods that avoid dissociation-induced stress artifacts.
Partition and barcode
Droplets, wells, or split-pool reactions attach cell and molecule identifiers.
- Measure
- Recovery · doublets
- Failure boundary
- Capture efficiency and doublet rate set a ceiling on usable data regardless of downstream depth.
Where the frontier moves
Cheaper, open partitioning chemistries like hydrogel-bead droplets.
Library and sequencing
Amplification and read allocation turn molecules into counts.
- Measure
- Reads/cell · saturation
- Failure boundary
- Sparse signal means deeper sequencing cannot recover molecules that were never captured.
Where the frontier moves
Sequencing cost continuing to fall per read while depth is allocated to purpose.
Inference
Quality control, integration, annotation, statistics, and validation turn matrices into claims.
- Measure
- Reproducible biological findings
- Failure boundary
- Batch effects and confounding can invalidate an otherwise well-executed experiment.
Where the frontier moves
Reference controls and standardized pipelines that make experiments comparable across labs.
A cell contains a finite, changing sample
Many measurements consume or perturb the cell, and transcripts are discrete molecules sampled with loss. Deeper sequencing cannot recover molecules that were never captured or states destroyed during preparation.
Cheap capture moves scarcity into samples and interpretation
Large atlases amplify the need for consented specimens, metadata, balanced experimental design, compute, reference standards, and orthogonal validation.
Preserve early
Stabilize tissue and benchmark dissociation bias before it silently shapes which cells are even measured.
Pool intelligently
Share sequencing runs across samples while retaining reliable per-sample identity through demultiplexing.
Sequence to purpose
Allocate reads based on the detection sensitivity and statistical power the question actually needs.
Use reference controls
Track sensitivity, doublets, ambient signal, and batch drift with standards, not assumptions.
Who is building what
Microfluidic droplet encapsulation, split-pool barcoding, in situ spatial transcriptomics, and single-cell proteomics resolve cellular heterogeneity. Search the record, or filter by profiling modality.
10x GenomicsChromium & XeniumGel Beads-in-emulsion (GEM) microfluidic partitioning for single-cell RNA-seq, coupled to Xenium in situ optical imaging of thousands of genes
- Reported evidence
- Standard platform behind thousands of published peer-reviewed single-cell atlases; Xenium routinely profiles millions of cells on tissue slides.
- Announced next step
- Lowering cost per cell while increasing spatial resolution to subcellular optical boundaries.
- Unresolved risk
- High proprietary consumable reagent pricing, instrument lock-in, and intellectual property litigation across microfluidic patents.
Parse BiosciencesEvercode Split-Pool IndexingInstrument-free single-cell barcoding using cells themselves as reaction compartments across multiple split-pool rounds of molecular indexing
- Reported evidence
- Commercial kits profiling up to 1 million cells per experiment without dedicated microfluidic hardware or specialized instruments.
- Announced next step
- Multi-million-cell population-scale drug screening and single-cell whole-genome CRISPR screening.
- Unresolved risk
- Multi-step pipetting protocol length, cell loss during repeated centrifugation washes, and fixation artifacts.
Bruker Spatial Biology / NanoStringCosMx Spatial Molecular ImagerHigh-plex optical barcoding with cyclic fluorescent hybridization, resolving 1,000+ RNA targets and proteins at single-cell and subcellular resolution
- Reported evidence
- Installed in academic and biopharma laboratories; demonstrated mapping of tumor microenvironment immune infiltrates on FFPE tissue.
- Announced next step
- Whole-transcriptome spatial mapping directly on standard clinical pathology glass slides.
- Unresolved risk
- Long instrument scan times per square millimeter of tissue and heavy computational burden of multi-terabyte optical image processing.
VizgenMERSCOPE PlatformMultiplexed Error-Robust Fluorescence in situ Hybridization (MERFISH) detecting hundreds of distinct RNA species with built-in error-correcting codes
- Reported evidence
- Widely deployed for neuroscience and oncology spatial mapping with near 100% transcript detection efficiency and exact photon localization.
- Announced next step
- Integration of spatial transcriptomics with simultaneous immunofluorescence protein quantification.
- Unresolved risk
- Probe library design constraints, tissue autofluorescence in difficult clinical specimens, and consumable cost per run.
Scale Biosciences (Scale Bio)Scale single-cell kitsArray-based combinatorial indexing for single-cell RNA, methylation, and multi-omics without microfluidic emulsion cartridges
- Reported evidence
- Commercial kits launched enabling 100k+ cell experiments with lower sequencing library preparation costs.
- Announced next step
- Ultra-high-throughput epigenetic profiling (single-cell ATAC and methylome) at single-cell resolution.
- Unresolved risk
- Sequencing cost dominates the total experiment cost as cell counts scale into the millions.
SeerProteograph Product SuiteEngineered multi-nanoparticle coronas that sample broad dynamic ranges of proteins from biological fluids, analyzed via mass spectrometry
- Reported evidence
- Commercial adoption for deep, unbiased plasma proteomics, quantifying thousands of proteins without antibody affinity reagents.
- Announced next step
- Extending nanoparticle protein sampling to single-cell and low-input tissue biopsies.
- Unresolved risk
- Mass spectrometry throughput bottlenecks and identifying low-abundance regulatory proteins in the presence of dominant albumin/globulin.
Human Cell Atlas (HCA)Global Reference AtlasInternational open-science consortium mapping all cell types, states, lineages, and spatial locations across healthy human tissues
- Reported evidence
- Published comprehensive organ-specific cell atlases covering lung, kidney, immune, nervous, and maternal-fetal tissues.
- Announced next step
- Complete first draft of the full human body cell atlas, integrating single-cell transcriptomic, spatial, and epigenetic datasets.
- Unresolved risk
- Data harmonization across diverse platforms, computational batch-effect correction, and ancestral representation across donor cohorts.
Broad InstituteDrop-seq & Perturb-seqFoundational methodology development combining single-cell RNA-seq with pooled CRISPR perturbations to map gene function at cellular scale
- Reported evidence
- Pioneered microfluidic Drop-seq and genome-scale Perturb-seq, mapping functional regulatory networks in human immune and cancer cells.
- Announced next step
- High-throughput foundational single-cell foundation models trained on billions of single-cell profiles.
- Unresolved risk
- Drop-out noise in sparse single-cell matrices confounding causal regulatory inference.
Cell throughput figures describe captured barcoded cells; useful biological information depends strictly on sequencing read depth per cell and drop-out rates for low-abundance transcripts.
The optimistic view, with conditions
Cell measurement becomes routine quality-controlled infrastructure
Costs will continue falling when open chemistries, standardized controls, sample multiplexing, and analysis pipelines make experiments comparable rather than merely large.
Open chemistries keep falling
Hydrogel-bead and other open platforms are pushing reagent cost well below proprietary comparators.
Allocate depth to purpose
Falling per-read prices matter less than matching depth to the statistical question being asked.
Standardize, don't just scale
Reference controls and shared pipelines are what make a large atlas comparable to another lab's.
What a cheap, trustworthy single-cell dataset actually needs
- Representative specimensPreservation and dissociation methods benchmarked against known bias.
- Matched sequencing depthReads allocated to the sensitivity and power the experiment actually requires.
- Reference controlsStandards that track sensitivity, doublets, ambient signal, and batch drift.
- Reliable multiplexingPooling that reduces cost without breaking sample identity.
- Reproducible pipelinesAnalysis that makes one lab's atlas comparable to another's.
Millions of cells still leave a coverage problem
The Human Cell Atlas portal lists about 70.9 million cells from 11,300 donors across 532 projects at this September 2026 snapshot. The unit “cell” conceals large differences in tissue, age, disease, ancestry, spatial context and molecular depth. The HyDrop example above is an ATAC assay of chromatin accessibility; its reagent price cannot be presented as a single-cell RNA-transcriptome price. Sequencing, tissue procurement, quality control and interpretation must share the denominator.
The next frontier asks how cells respond, rather than only what state they occupy. Arc Institute’s 2025 virtual-cell atlas includes Tahoe-100M, a 100-million-cell perturbation dataset across 50 cancer-cell models and 60,000 drug–cell interactions. Its challenge benchmark adds roughly 300,000 cells under targeted gene perturbations. Large cell counts are evidence of throughput; coverage of normal human tissues and predictive performance on held-out perturbations are separate quality tests.
A published full-length single-cell RNA study reported about $3 per cell for a 960-cell sequencing run versus about $60 for a comparator, on its specific workflow. That is a method comparison, not a universal current quote or a total atlas cost. An honest cost curve must state assay type, reads and molecules per cell, failed cells, labour and computation; spatial methods also need a cost per sampled tissue area.
The cost curve needs a biological denominator
At the atlas scale, a captured droplet is not automatically an analyzable cell. Human Cell Atlas data aggregate projects with different tissues and assays; the 70.9-million-cell portal total is a throughput snapshot, not a single standardized price series. The fixed biological hurdle is capturing enough molecules from each cell to distinguish rare populations and low-expression genes above dropout and ambient noise. That threshold varies by question, so there is no universal “RNA molecules per cell” cost floor.
For an honest comparison, report total cost per retained, interpretable cell at stated median reads, genes detected, doublet rate, tissue procurement and analysis. Spatial assays need a second unit, usable cells per tissue area at stated resolution. Published method comparisons, such as the full-length RNA workflow discussed above, establish a specific experimental boundary; list prices for reads alone do not include acquisition or laboratory labour. The public evidence does not yet support a like-for-like annual cost curve across RNA, ATAC and spatial profiling.
Sources, method, and boundaries
Cost examples retain their explicit exclusions and are not generalized across platforms. The comparison framework follows measured workflow, sensitivity, recovery, and sequencing-efficiency dimensions. The interactive calculator combines the HyDrop v2 reagent figure with public sequencing-provider pricing as an illustration of how capture cost and sequencing depth trade off, not a reproduction of any study's own cost accounting.
- Doublet
- A droplet or well erroneously containing two cells, counted and later filtered as one contaminated sample.
- Ambient RNA
- Background transcripts from lysed cells that contaminate droplets and must be computationally corrected.
- Saturation
- The point at which additional sequencing reads stop revealing new molecules per cell.
