Single Cell RNA Sequencing Protocol: A Step-by-Step Experimental Workflow

A single cell RNA sequencing protocol is not a reagent list. It is a set of decisions made before the sample exists, acceptance criteria at each bench stage, and controls chosen to make the resulting data interpretable. Almost every irrecoverable failure in single-cell RNA sequencing happens before sequencing — in a design choice that cannot be corrected computationally, or in a sample-preparation step whose damage is indistinguishable from biology once it is in the data.

This guide assumes you already understand cell barcodes, UMIs, and why droplet partitioning works. If you do not, read the conceptual companion first — Single Cell RNA Sequencing Workflow: A Researcher's Primer — and come back. What follows is strictly executional.

What a single cell RNA sequencing protocol must specify before you book instrument time

Most published single-cell methods sections describe what was done. A working protocol has to describe what will be decided, and on what evidence. A complete single cell RNA sequencing protocol specifies eight things. If any is unspecified when you book the instrument, that is the thing that will go wrong:

• The unit of replication — how many biological replicates per condition, decided before cell numbers.
• The batch structure — which samples share a run, and how conditions are distributed across runs.
• The cells-versus-depth split — target cell recovery and reads per cell, at a fixed budget.
• The material state — fresh, cryopreserved, fixed, or nuclei, decided against known biases rather than convenience.
• The dissociation protocol — enzyme, temperature, duration, and the artifact it will introduce.
• Acceptance criteria — the viability, concentration, and debris thresholds that authorize loading, written down in advance.
• The controls — which technical and biological controls run, and what each one detects.
• The stop conditions — what result at each checkpoint means abort and re-prepare rather than proceed.

Single cell RNA sequencing experimental design: five decisions before the sample exists

Single cell RNA sequencing experimental design is unusual in that the most consequential decisions are made when there is nothing yet to measure. Each of the following has a default that is wrong for most studies, and a correct answer that depends on the question.

1. How many biological replicates — not how many cells

Cells within one animal, donor, or culture are not independent observations of a treatment effect. Testing them as though they were inflates significance dramatically: methods that ignore variation between biological replicates are biased and produce false discoveries, and the fix is to account for replicate structure rather than to add cells.^3,4 Aggregating cells to a per-sample pseudobulk profile and testing across samples is the robust default for condition comparisons.^5,6

2. Which samples get processed together — batching, blocking, and multiplexing

A batch is any group of samples that went through a processing step together, sharing that step's reagent lot, operator, and conditions. Several steps each create their own batch: the day a sample was dissociated, the chip it was partitioned on, the library prep, the sequencing run. Batch effects in single-cell studies are substantial and are best handled by design rather than by correction after the fact.¹

Processing a study across several days is not itself the problem. The problem is letting a batch line up perfectly with condition. If every control sample is dissociated on Monday and every treated sample on Tuesday, any difference between them has two equally good explanations — the treatment, or the day — and nothing in the data can choose between them.

The critical point, and the one most easily missed: pooling samples onto one chip does not remove every batch. It removes the downstream ones only. Cells dissociated on different days already carry that difference before they reach the instrument.

Dissociating everything in one session pulls against a second requirement, and the two are easy to confuse because both get described as happening “on the same day.” For any individual sample, the interval from collection to loading should be as short as possible, because stress signature and RNA degradation both accumulate with elapsed time. Across samples, a single shared dissociation session is what removes dissociation as a batch. Those two goals are compatible only when every sample is available at once, so the resolution depends on how long your collection window is.

That leaves two tools, used together rather than instead of each other. Blocking puts both conditions into every batch at the steps pooling cannot fix — a study dissociated over four days handles some control and some treated material each day, so batch still exists but is no longer the same variable as condition. Multiplexing pools samples into one run and separates them afterwards, either from natural genetic variation between donors⁷ or from oligo-tagged antibodies that label each sample before pooling.⁸

3. Cell number versus sequencing depth — a fixed-budget trade

At a fixed cost these trade directly. More cells at shallower depth resolves composition and rare populations; fewer cells at greater depth resolves expression differences within a defined cell type. Sensitivity and cost per cell differ several-fold between protocols at matched depth, so this is a decision to make against published benchmarks rather than a platform default.^9,10

4. Material state: fresh, cryopreserved, fixed, or nuclei

This is usually decided by logistics, which is why it so often goes wrong. A systematic comparison across matched mouse kidney samples found that cryopreservation of dissociated cells caused a major loss of epithelial cell types, while methanol fixation preserved cellular composition but suffered from ambient RNA leakage.² Those are different failure modes with different consequences: one distorts your composition estimate, the other adds background to every cell.

Nuclei deserve separate comment, because choosing them changes what you measure rather than just how you store it. Single-nucleus protocols release nuclei by mechanical homogenization, avoiding warm proteolysis entirely — which is why they are standard for frozen tissue, brain, muscle, and adipose. The cost is that the nuclear transcriptome is not the cellular transcriptome: cytoplasmic mRNAs are underrepresented and unspliced transcripts enriched, and matched comparisons recover different cell-type proportions from the same tissue.^2,10

Choose nuclei when the tissue resists gentle dissociation (brain, skeletal and cardiac muscle, adipose), the material is already frozen or archived, your collection window is long, or a pilot shows the dissociation stress signature dominates. Choose cells when you need cytoplasmic or mitochondrial transcripts, or when you are co-measuring surface protein, which requires an intact membrane.

A caveat on the evidence: the strongest published comparison covers one tissue in one species. Treat the specific findings as a reason to test your own material rather than as constants that transfer to every tissue.

5. Which controls will run

Most single-cell controls are only informative if they were built into the run. A species-mixing control cannot be added after the fact; a batch anchor is meaningless unless the same material was present in every run; a dissociation-artifact comparison requires that one sample was deliberately split. See the controls section below for what each detects and when it has to be committed.

Single cell RNA sequencing sample preparation: dissociation, storage, and the suspension you load

Single cell RNA sequencing sample preparation is where the majority of recoverable quality is won or lost, and it is the stage most often treated as a preliminary rather than as part of the experiment.

Dissociation: temperature is a variable, not a detail

Warm proteolytic dissociation at 37 °C induces a transcriptional stress response — immediate-early genes and heat-shock genes — in a subset of cells, strong enough that it can be mistaken for a biological cell state.¹¹ Cold-active proteases from psychrophilic organisms used at 4–6 °C substantially reduce this signature, demonstrated first in kidney¹² and subsequently in solid tumors, where collagenase-associated stress responses are conserved across tumor types.¹³ Digestion on ice has been independently confirmed to avoid the 37 °C stress response.²

Three parameters should be fixed in writing and held constant across every sample in a study: enzyme and concentration, temperature, and duration. Varying any of them between conditions creates an artifact that tracks with condition.

What the suspension has to satisfy before loading

Write these thresholds down before the sample arrives, so that the decision to proceed is not made under time pressure with a dissociated sample in hand.

Record the actual measured values, not just pass or fail. A run that later looks odd is far easier to diagnose when you can see that viability was 81% rather than “acceptable.”

The single cell RNA sequencing workflow as eight go/no-go gates

The single cell RNA sequencing workflow below is expressed as checkpoints rather than as narrative. Each gate has an input, an action, a criterion, and a stop condition. Gates marked as stops are points where proceeding with a failing sample wastes the remaining reagent and instrument cost.

Gate 1 — Tissue receipt and dissociation

Digest to a single-cell suspension using the pre-specified enzyme, temperature, and duration. Record elapsed time from excision. Criterion: suspension is visibly homogeneous with no residual tissue fragments. Stop if digestion required extension beyond the specified duration — note it as a protocol deviation rather than silently absorbing it, because extended digestion changes the artifact profile for that sample only.

Gate 2 — Filtration, washing, and depletion

Pass through a strainer sized to the platform requirement. Remove dead cells and, for blood-containing tissue, erythrocytes. Nuclease treatment here reduces DNA-mediated clumping. Criterion: no visible clumps; supernatant clears on wash. Stop if repeated washing fails to disperse the suspension — loading a clumpy prep guarantees an uninterpretable multiplet rate.

Gate 3 — Count and viability

Determine concentration and per-cell viability on the final suspension immediately before loading. Criterion: both within the pre-specified windows recorded in the protocol. Stop if viability is below threshold. This is the highest-value stop in the entire workflow: everything downstream is expensive and nothing downstream repairs a low-viability input.

Gate 4 — Partitioning and barcoding

Load at the concentration corresponding to your target recovery and accepted multiplet rate. If multiplexing, samples are pooled at this point in known proportions. Criterion: emulsion appearance normal and uniform per platform guidance; loading volume and concentration recorded. Note that the multiplet rate is now fixed by the loading choice — it can be measured and modeled afterwards, but not reduced.

Gate 5 — Reverse transcription and cDNA amplification

Barcoded cDNA is generated inside the compartments, then droplets are broken and material pooled for amplification. Criterion: amplified cDNA yield and trace within expected range for the input cell number. Investigate if yield is low: this points either to poor capture (check RT conditions and RNA integrity) or to fewer recovered cells than intended.

Gate 6 — Cleanup, size selection, and library construction

Bead-based size selection removes primers, adapter dimers, and off-target short products between steps; the library is then fragmented, adapter-ligated, and indexed. Criterion: library trace shows expected fragment distribution with no adapter-dimer peak. Stop if a dimer peak persists — it consumes sequencing capacity and reduces usable reads per cell, which is the resource this entire protocol has been budgeting.

Gate 7 — Sequencing

Sequence with the platform-specified asymmetric read configuration and the depth per cell fixed at design time. Criterion: barcode read quality and per-cell read depth meet target. Check early: a shallow pilot lane before committing full depth is inexpensive relative to a full run built on a poor library.

Gate 8 — Cell calling and pre-analysis QC

Distinguish genuine cell-containing barcodes from empty droplets, then filter on quality metrics before any biological interpretation. Criterion: cell calling should test each barcode against the ambient profile rather than applying a simple count cutoff, which discards genuinely RNA-poor cell types.¹⁴ Filter thresholds for mitochondrial fraction and gene count should be set from the observed distribution, not from defaults.^15,16,20 Stop and reconsider if recovered cell number is far below target, or if a stress or ambient signature dominates — interpreting such a dataset produces confident conclusions about your protocol rather than your biology.

Single cell RNA sequencing controls and what each one detects

Single cell RNA sequencing controls are discussed far less than analysis methods, which is backwards: an uncontrolled run generates numbers that look identical to a controlled one and cannot be audited afterwards. The controls below each detect a specific, named failure. Most must be committed at design time.

Species-mixing control — measures your actual multiplet rate

Mix human and mouse cells at a known ratio and run them as a single sample. Because reads map unambiguously to one genome or the other, any barcode receiving both must be a multiplet — giving a direct empirical measurement of the multiplet rate at your loading concentration, on your instrument, in your hands. This design has been used since the founding droplet papers to characterize partitioning performance.^17,18 Run it once when commissioning a platform or protocol change, not every run. What it does not tell you: the within-sample multiplet rate for cells of the same species, which is inferred from the cross-species rate rather than measured.

Batch anchor (bridge) control — detects technical drift between runs

Include an aliquot of the same cryopreserved cell line in every run of a study. Because the input material is identical, any difference in its profile between runs is technical. This converts batch effect from an assumption into a measurement, and gives integration methods a known-invariant reference to align against. Requirement: one large, homogeneous, well-characterized batch of frozen aliquots prepared at the start of the study. Splitting a fresh culture each time defeats the purpose, because culture passage becomes a new variable.

Dissociation-artifact control — quantifies your own stress signature

Split one representative sample and process the halves by two protocols — warm versus cold dissociation, or cells versus nuclei from matched tissue. The genes that differ between halves are your artifact, measured in your tissue rather than assumed from the literature. This is the only way to know whether the published stress signatures apply to your system and at what magnitude.^2,11

Multiplexing as a doublet control

When samples are pooled and demultiplexed, any barcode carrying two sample identities is a multiplet by observation rather than by inference. Genetic demultiplexing uses natural variation between donors,⁷ while oligo-tagged antibody hashing works for any samples including genetically identical ones.⁸ Limitation: only cross-sample multiplets are caught. Two cells from the same sample in one droplet remain invisible to this control and need computational detection.

Ambient RNA reference — already in your data, if you keep it

The empty droplets in your own run are a direct sample of the ambient RNA profile — a built-in negative control, provided the pipeline does not discard those barcodes before you look. Statistical cell calling works precisely by testing each barcode against this ambient profile.¹⁴ Retain and inspect the low-count barcode fraction; a pipeline that silently drops it removes your only measurement of the background being added to every cell.

Biological replicates — the control for a condition comparison

No technical control substitutes for independent biological replicates. Differential expression methods that fail to account for between-replicate variation are biased toward false discoveries regardless of how many cells were profiled.^3,4 Aggregating to per-sample pseudobulk and testing across samples is the robust default, and method benchmarking supports it over cell-level tests.^5,6,19

Orthogonal bulk arm — independent check on the aggregate

Bulk RNA-seq on matched material gives an independent measurement to compare against your pseudobulk aggregate. Systematic disagreement is informative: it flags cell types lost during dissociation, since a population absent from the suspension is absent from the single-cell aggregate but still represented in bulk from intact tissue.²

What no control will fix

Two failures are outside the reach of any control. First, a cell type that does not survive dissociation is simply absent — it produces no barcode, so no QC metric flags its absence and no correction restores it. The orthogonal bulk arm is the only way to suspect it. Second, batch fully confounded with condition cannot be rescued; integration can align datasets but cannot recover a comparison that was never separable. Both are design failures, which is why the design section comes first.

Troubleshooting a single-cell run by symptom

Pre-run checklist: sign this off before the instrument is booked

Design, fixed in writing

• Number of biological replicates per condition stated, with cell count treated separately
• Batch assignment written out for every step that creates one — dissociation day, chip run, library prep, sequencing run — with conditions distributed across each rather than stacked
• Multiplexing strategy chosen (genetic, hashing, or none) with justification
• Target cell recovery and reads per cell specified, derived from the rarest population of interest
• Material state decided (fresh, cryopreserved, fixed, nuclei) against known biases for the endpoint
• Dissociation enzyme, temperature, and duration fixed and identical across all samples

Controls committed

• Species-mixing control run at commissioning, multiplet rate on record for this loading concentration
• Batch anchor aliquots prepared as one homogeneous frozen batch, sufficient for every planned run
• Dissociation-artifact split scheduled on a representative sample
• Orthogonal bulk arm planned on matched material
• Pipeline configured to retain low-count barcodes for ambient profiling

Acceptance criteria recorded in advance

• Viability threshold, concentration window, and debris standard written down before the sample arrives
• Stop conditions defined at each gate, with a named person authorized to call the stop
• Deviation log ready, so extended digestion or repeated washing is recorded rather than absorbed
• Measured values — not pass/fail — captured at every checkpoint

Protocol-stage reagents: what the BioHippo catalog covers, and what it does not

Control cell lines

The species-mixing control needs a human and a mouse line profiled together. HEK293 Cells and NIH-3T3 Cells are the conventional pairing for this purpose, being well-characterized, easy to culture, and genomically unambiguous between species. For a batch anchor, a suspension line avoids the trypsinization variable at each thaw: Jurkat Cells and K562 cells are both widely used and heavily characterized in public single-cell datasets, which helps when you need a reference profile to compare your anchor against. Browse the full range of cell lines and Cytion authenticated lines.

One caveat worth raising with your core facility: cell line identity should be confirmed by STR profiling before a line becomes the invariant reference for a multi-run study. A misidentified anchor silently invalidates every batch comparison built on it.

Nuclease treatment

Recombinant Deoxyribonuclease I (DNase I, RNase-free) is a recombinant, RNase-free DNase I supplied at 2 U/µL with reaction buffer; a GMP-grade version is also available. Being explicit about scope: the supplier's stated applications are genomic DNA removal before RNA extraction or reverse transcription, template removal after in vitro transcription, and rRNA removal during library construction — nucleic-acid workflow steps. DNase I is also routinely included in tissue dissociation buffers to limit DNA-mediated clumping, but that is general practice for the enzyme class rather than a validated application of this particular product, and you should confirm suitability for that use yourself.

Nuclei isolation

Decision 4 recommends nuclei for long collection windows and hard-to-dissociate tissue, so it is worth being explicit about what that prep needs. It is not a single product: nuclei are released with a hypotonic, detergent-containing lysis buffer plus mechanical disruption in a homogenizer, then cleaned up through a density gradient and strainers, with a carrier protein to stop nuclei adhering to plastic and a nuclear stain to confirm no intact cells remain. Of that list, the catalog covers one component.

That component matters, though. Lysing the plasma membrane releases cytoplasmic ribonucleases directly into the buffer holding your nuclei, so an RNase inhibitor is standard in nuclei isolation buffers rather than optional. RNaseOFF Ribonuclease Inhibitor is a protein inhibitor of RNase A, B, and C, described by the supplier as resistant to oxidation and stable at DTT concentrations below 1 mM, with no downstream polymerase inhibition. Two pack sizes are listed under the same product title (catalogue references G138 and G591). Being precise about scope: the supplier's stated application is as an additive in PCR and RT-PCR, not nuclei isolation specifically — RNase inhibitors are standard practice in nuclei buffers for the enzyme class, but confirm suitability for that use yourself.

Everything else in a nuclei prep — lysis buffer, homogenizer, gradient medium, carrier protein, strainers, nuclear stain — comes from elsewhere. If you are planning a single-nucleus study, source those first and treat the RNase inhibitor as a line item to add rather than a starting point.

Cleanup and size selection

Hieff NGS™ DNA Selection Beads are SPRI-principle magnetic beads for the DNA purification and size-selection steps at Gate 6, stated by the supplier to be compatible with major DNA and RNA library prep kits. Two specifics: the standard formulation recovers fragments of roughly 150 bp and longer, and the buffer is formulated for DNA rather than RNA.

Orthogonal bulk arm and validation

For the matched bulk comparison described in the controls section, Hieff™ Cell/Tissue Total RNA Kit covers extraction and Hieff NGS™ Ultima Dual-mode RNA Library Prep Kit covers strand-specific library construction from total RNA, requiring a separately purchased platform primer mix and an upstream mRNA-purification or rRNA-depletion step. For RT-qPCR validation of markers your analysis nominates, Hifair™ Super Reverse Transcriptase, Glycerol-free (600 U/µL) is an engineered M-MLV (H−) derivative operating up to 70 °C. It is not a validated single-cell chemistry and will not substitute for a platform kit. Related lines: Molecular Biology Enzymes.

Frequently asked questions

Do I need to run a species-mixing control every time?

No. It characterizes the multiplet rate for a given loading concentration on a given instrument, so it is a commissioning control — run it when you adopt a platform, change loading concentration, or change dissociation protocol substantially. Repeating it per run spends cells and reagent on a number you already know.

Can I add controls after the run if the data looks strange?

Almost none of them. A batch anchor is only interpretable if the same material was in the earlier runs; a dissociation-artifact split requires that a sample was divided at prep; a species-mixing control cannot be retrofitted at all. The one genuine exception is the ambient RNA profile, which is already in your data as the empty-droplet fraction — provided your pipeline did not discard those barcodes.¹⁴

How do I set viability and mitochondrial thresholds?

Viability is a pre-load bench threshold and should come from your platform's specification, written down before the sample arrives. The mitochondrial-fraction filter is an analysis threshold and should be set from the observed distribution in your own data rather than from a copied default, because the appropriate cutoff varies by tissue and by cell type — cardiomyocytes legitimately carry a high mitochondrial fraction.^15,16

Is it better to overload and remove doublets computationally?

Only if you are multiplexing. With pooled, demultiplexed samples, cross-sample multiplets become identifiable, so deliberate overloading recovers more usable single cells per run.⁸ Without multiplexing, overloading raises a multiplet rate that only computational inference can address, and same-sample doublets remain the hardest case.

My tissue will not dissociate without harsh treatment. What are my options?

Move to nuclei rather than escalating the digestion. Single-nucleus protocols release nuclei mechanically and avoid warm proteolysis entirely, which is why they are standard for brain, muscle, adipose, and frozen material. Accept that you are measuring a different compartment: cytoplasmic transcripts are underrepresented and cell-type proportions differ from matched single-cell data.^2,10

Can cryopreservation let me batch samples collected over months?

It is the standard approach to that logistics problem, but validate it before committing the study. Cryopreservation of dissociated cells has been shown to cause major loss of epithelial cell types in matched comparisons, while methanol fixation preserved composition but increased ambient RNA leakage.² If your endpoint is cell-type proportions, run a fresh-versus-stored arm on one sample first.

How many replicates is enough for single-cell RNA sequencing?

There is no universal number, but the framing matters more than the count: replicates are samples, not cells, and adding cells does not increase power for a condition comparison. Published analyses of this error found that ignoring between-replicate variation drives false discoveries irrespective of cell number.^3,4 Three or more biological replicates per condition is a common practical floor; power depends on the effect size and the cell type's abundance.

Should I use spike-ins?

Synthetic RNA spike-ins are practical in plate-based full-length protocols, where a known quantity can be added to each well and used to assess capture efficiency and technical noise. They are not compatible with most high-throughput droplet chemistries, where the ambient profile and the empty-droplet fraction serve a comparable diagnostic role instead.

Sourcing the controls and reagents your protocol names

Control cell lines, nuclease treatment, cleanup beads, and the bulk validation arm are all things you can consolidate into one order. The partitioning chemistry, hashing panels, and cold-active proteases are not — and this guide names which is which. Tell a BioHippo technical specialist your tissue, platform, and control plan, and we will point you to what we stock and be direct about what we do not. Talk to a specialist or request a quote.

References

1. Tung P-Y, Blischak JD, Hsiao CJ, et al. Batch effects and the effective design of single-cell gene expression studies. Sci Rep. 2017;7:39921. PMID 28045081
2. Denisenko E, Guo BB, Jones M, et al. Systematic assessment of tissue dissociation and storage biases in single-cell and single-nucleus RNA-seq workflows. Genome Biol. 2020;21(1):130. PMID 32487174
3. Squair JW, Gautier M, Kathe C, et al. Confronting false discoveries in single-cell differential expression. Nat Commun. 2021;12(1):5692. PMID 34584091
4. Zimmerman KD, Espeland MA, Langefeld CD. A practical solution to pseudoreplication bias in single-cell studies. Nat Commun. 2021;12(1):738. PMID 33531494
5. Murphy AE, Skene NG. A balanced measure shows superior performance of pseudobulk methods in single-cell RNA-sequencing analysis. Nat Commun. 2022;13(1):7851. PMID 36550119
6. Crowell HL, Soneson C, Germain P-L, et al. muscat detects subpopulation-specific state transitions from multi-sample multi-condition single-cell transcriptomics data. Nat Commun. 2020;11(1):6077. PMID 33257685
7. Kang HM, Subramaniam M, Targ S, et al. Multiplexed droplet single-cell RNA-sequencing using natural genetic variation. Nat Biotechnol. 2018;36(1):89–94. PMID 29227470
8. Stoeckius M, Zheng S, Houck-Loomis B, et al. Cell Hashing with barcoded antibodies enables multiplexing and doublet detection for single cell genomics. Genome Biol. 2018;19(1):224. PMID 30567574
9. Ziegenhain C, Vieth B, Parekh S, et al. Comparative analysis of single-cell RNA sequencing methods. Mol Cell. 2017;65(4):631–643.e4. PMID 28212749
10. Ding J, Adiconis X, Simmons SK, et al. Systematic comparison of single-cell and single-nucleus RNA-sequencing methods. Nat Biotechnol. 2020;38(6):737–746. PMID 32341560
11. van den Brink SC, Sage F, Vértesy Á, et al. Single-cell sequencing reveals dissociation-induced gene expression in tissue subpopulations. Nat Methods. 2017;14(10):935–936. PMID 28960196
12. Adam M, Potter AS, Potter SS. Psychrophilic proteases dramatically reduce single-cell RNA-seq artifacts: a molecular atlas of kidney development. Development. 2017;144(19):3625–3632. PMID 28851704
13. O'Flanagan CH, Campbell KR, Zhang AW, et al. Dissociation of solid tumor tissues with cold active protease for single-cell RNA-seq minimizes conserved collagenase-associated stress responses. Genome Biol. 2019;20(1):210. PMID 31623682
14. Lun ATL, Riesenfeld S, Andrews T, et al. EmptyDrops: distinguishing cells from empty droplets in droplet-based single-cell RNA sequencing data. Genome Biol. 2019;20(1):63. PMID 30902100
15. Ilicic T, Kim JK, Kolodziejczyk AA, et al. Classification of low quality cells from single-cell RNA-seq data. Genome Biol. 2016;17:29. PMID 26887813
16. Luecken MD, Theis FJ. Current best practices in single-cell RNA-seq analysis: a tutorial. Mol Syst Biol. 2019;15(6):e8746. PMID 31217225
17. Macosko EZ, Basu A, Satija R, et al. Highly parallel genome-wide expression profiling of individual cells using nanoliter droplets. Cell. 2015;161(5):1202–1214. PMID 26000488
18. Zheng GXY, Terry JM, Belgrader P, et al. Massively parallel digital transcriptional profiling of single cells. Nat Commun. 2017;8:14049. PMID 28091601
19. Soneson C, Robinson MD. Bias, robustness and scalability in single-cell differential expression analysis. Nat Methods. 2018;15(4):255–261. PMID 29481549
20. Heumos L, Schaar AC, Lance C, et al. Best practices for single-cell analysis across modalities. Nat Rev Genet. 2023;24(8):550–572. PMID 37002403

This guide summarizes published single-cell RNA sequencing methodology and is provided for research reference. It is not a platform-specific protocol; follow your instrument vendor's current documentation for loading concentrations, read configurations, buffer compositions, and viability specifications. Performance findings cited are from the referenced studies, not from BioHippo testing. BioHippo does not supply single-cell partitioning instruments, barcoded bead kits, hashing antibody panels, cold-active dissociation proteases, nuclei isolation kits or lysis buffers, or complete scRNA-seq library chemistries; the catalog products referenced are general cell-culture, molecular-biology, and NGS reagents and are not validated for single-cell library preparation. Cell line identity should be independently confirmed by STR profiling before use as a reference control. Confirm specifications and intended-use statements on each product page before purchase. All products are for research use only. Reference metadata verified via PubMed.