Before You Outsource scRNA-seq: 5 Questions About Your Study Design

For research use only. Not intended for clinical diagnosis, treatment, or individual health assessment.
A technically successful scRNA-seq project can still leave its main research question unanswered. Good libraries and abundant reads cannot compensate for unsuitable controls, unrepresentative sampling, or an analysis plan that ignores biological replication. Before commissioning sequencing, define what evidence would support your hypothesis and which decisions must be settled before specimens enter the laboratory.
The five questions below connect study value, workflow selection, preparation review, tissue sampling, and analysis scope. They are particularly useful when biotech, CRO, and academic teams are turning an initial idea into a project proposal. For general supplier and delivery considerations, see outsourcing scRNA-seq. For resource allocation, see scRNA-seq budget planning.
Does My Study Add Value Beyond Published Cell Atlases?
An existing atlas does not automatically make a new study redundant. It may describe baseline cell identities without covering your intervention, disease state, anatomical region, or sampling time. Equally, collecting another descriptive dataset may add little if suitable public data already answer the question. Start by reviewing single-cell and spatial omics atlas resources, then identify the evidence that is still missing.
Compare species, tissue compartment, donor characteristics, preservation, chemistry, and annotation depth. An atlas can support reference mapping and hypothesis generation. It is not automatically a matched control for a newly collected treatment cohort. Differences in collection and preparation can become entangled with biological differences, even after computational integration [1,4].
Consider a hypothetical study of a kinase inhibitor in a tumor model. A public tumor atlas may help identify expected immune and stromal populations. It may not provide an appropriate untreated baseline for that model, dose, and collection window. Concurrent treatment and vehicle specimens, balanced across processing batches, could address this gap. If the objective is simply to check whether a candidate gene is expressed in a known cell population, reanalyzing existing data may be a useful first step instead.
| Research Need | What Existing Data May Provide | Missing Evidence | Possible New Study Contribution |
|---|---|---|---|
| Cell identity | Reference markers and broad cell classes. | A state absent from the reference. | Profile relevant specimens and validate candidate state markers. |
| Treatment response | Background expression and hypotheses. | Appropriate treatment and control comparisons. | Measure intervention-associated changes across independent specimens. |
| Time-dependent changes | A snapshot of one condition. | Defined collection times and comparable controls. | Compare planned time points; validate inferred transitions independently. |
| Regional differences | Profiles from incompletely documented tissue regions. | Region-specific sampling provenance. | Analyze documented regions separately when regional comparison matters. |
More cells do not replace more independent biological specimens. An illustrative atlas containing 50,000 cells from two donors still has only two independent donor-level observations. Its statistical degrees of freedom depend on the model, rather than simply equaling the donor count. Additional cells improve within-specimen resolution, while independent donors support population-level comparisons [2].
There is no universal sample number that guarantees power. Expected effect size, between-donor variability, target-cell abundance, and the planned statistical test matter. Pilot data can support sample-size assessment; published methods such as scPS explicitly consider within-sample cell correlations [3]. Use such calculations with their assumptions, rather than copying a sample count from an unrelated publication.
Before requesting a quote, write one sentence defining the primary comparison and one defining the missing evidence. If neither is clear, narrow the question before adding assays. Research value should come from an interpretable comparison, not from promising novelty or publication.
Which Workflow Fits My Samples and Research Question?
Choose the required output before choosing an instrument. Cell-type discovery, paired immune receptors, transcript isoforms, and spatial location require different evidence. A workflow that produces excellent expression profiles may still be unsuitable for a receptor or isoform endpoint.
Standard 3' gene expression profiling can characterize immune populations and activation-associated expression. It does not routinely provide reliable paired TCR/BCR sequences. Within 10x workflows, paired receptor analysis generally uses compatible 5' chemistry plus dedicated V(D)J libraries. BD Rhapsody also offers receptor workflows alongside expression profiling; paired receptor recovery is not exclusive to 5' expression assays across all platforms [7]. Compare outputs in 3' versus 5' scRNA-seq.
For conventional expression projects, evaluate the specific chemistry, input format, target population, and laboratory preparation. The Next GEM, GEM-X, and BD Rhapsody comparison explains these choices. Salcher and colleagues compared workflows in paired human prostate cancer specimens [5]. Their findings help frame platform questions, but do not establish a universal preference for low-input samples or fragile granulocytes.
Frozen tissue may favor snRNA-seq when intact-cell recovery is impractical. Properly cryopreserved cell suspensions are a different input and may support scRNA-seq after validated thawing. Cell and nuclear profiles have different representation and RNA-content biases [4]. See nuclei isolation versus tissue dissociation before selecting a fallback.
FFPE requires a compatible fixed-sample workflow, rather than assuming that a conventional live-cell protocol will work. Confirm species, probe coverage, tissue preparation, and input requirements against the exact protocol. Long-read transcript profiling may add value for isoforms; ATAC measures accessibility, not direct regulatory causality. Spatial RNA measurements add location, but are not interchangeable with dissociated expression data. Manufacturer capabilities should be confirmed against the service workflow actually proposed.
What Counts as an Acceptable Preparation—and What Happens If It Fails?
"High-quality sample" is an incomplete acceptance criterion. Agree on what will be reviewed, when the laboratory will pause, and who authorizes further work. Distinguish incoming-material inspection, post-preparation assessment, and library/data review. Each stage addresses different risks.
Incoming review checks identity, quantity, preservation history, and transport condition. Preparation review examines intact singlets or nuclei, concentration, recovery, debris, aggregates, and viability where applicable. Library and data review consider library profiles, sequencing quality, cell calling, background, multiplets, and target-population representation. No individual measurement establishes that a specimen adequately represents the intended tissue [1,4].
For example, a hypothetical suspension at 65%–75% viability merits a documented review; this range is not a general acceptance threshold or a prediction of success. Likewise, a project-specific goal above 80% should not be treated as a universal boundary. The relevant question is whether the remaining intact cells include the population needed for the study. General preparation protocols require sample-specific optimization [8].
| Finding | Research Risk | Decision to Confirm |
|---|---|---|
| Reduced viability or abundant debris | Background RNA and selective population loss. | Review cleanup, an alternative input, or risk-authorized continuation; record expected losses. |
| Microscopically confirmed cell aggregates | Multiplets, loading problems, and biased recovery. | Evaluate additional processing against target-cell size and fragility; pause if unsuitable. |
| Recovery below the planned target | Fewer opportunities to observe rare populations. | Revise the target, use reserve material, or reconsider feasibility before library preparation. |
| Damaged nuclei | RNA leakage, background, and reduced complexity. | Review extraction conditions or re-extract reserve tissue before proceeding. |
A hypothetical change from 8,000 to 3,000 recovered cells should trigger review of the endpoint, not just a lower loading target. Cleanup can itself remove relevant cells. Computational background correction cannot restore missing populations or RNA. Record the chosen action, additional costs, schedule changes, and client authorization. Detailed laboratory procedures belong in the validated preparation plan, including when using suspension preparation services.
How Much Tissue Should I Collect, and Which Region Should I Sample?
A suspension with 90% viability may still be poorly suited to the study if it comes from the wrong region. Tissue amount and tissue identity are separate decisions. Sampling a tumor edge, central area, or adjacent tissue can change the populations available for analysis, regardless of library quality.
Figure 2. Conceptual comparison of one region, separately profiled regions, and regions pooled before dissociation. The designs retain different levels of regional information.
Published evidence supports taking regional variation seriously. Langerud and colleagues analyzed multiregional colorectal cancer transcriptomes and demonstrated variation between tumor regions [6]. This was a bulk transcriptomic study, not a benchmark establishing an optimal number of regions for scRNA-seq. Its relevance here is the sampling principle: one fragment may not represent the whole lesion.
- One documented region: Appropriate when the endpoint concerns a defined compartment. It cannot establish whole-organ representativeness without additional evidence.
- Regions profiled separately: Retains regional identity and supports region-level comparisons. Regions from one donor remain correlated; they are not independent donor replicates. Additional libraries increase costs, although not necessarily in direct proportion.
- Regions pooled before dissociation: May broaden coverage under budget constraints, but cell-level regional origin is lost without a validated identity-preserving strategy. Unequal tissue contributions or preparation losses can dominate the pooled profile.
Region labels identify where material was collected; they do not restore precise cell coordinates after dissociation. If the endpoint depends on neighborhood relationships or continuous spatial gradients, evaluate an appropriate spatial assay or independent tissue imaging.
| Endpoint | Sampling Consideration | Separate or Pool? | Potentially Missed Information | Record or Validate |
|---|---|---|---|---|
| Tumor–immune boundary | Document the interface and comparison region. | Separate if regional contrasts matter. | Central or distant-region states. | Orientation, region photographs, adjacent histology. |
| Broader organ coverage | Sample relevant anatomical compartments. | Keep compartments separate when identity matters. | Local niches hidden by pooling. | Location and contribution of each fragment. |
| Treatment-associated changes | Use comparable regions across groups; include injury zones if relevant. | Retain donor identity and planned paired comparisons. | Damage-associated states excluded by selection. | Exposure, collection timing, tissue condition. |
| Fibrotic remodeling | Consider lesion and less affected tissue. | Separate regions or evaluate nuclear preparation as appropriate. | Cells lost through difficult dissociation. | Histology, preparation conditions, target-cell recovery. |
Do not routinely exclude necrosis, fibrosis, or tissue borders. Exclusion may be sensible for some endpoints, but harmful when injury, resistance, or stromal remodeling is the question. Record exclusions and why they were made. Dissociation can further alter representation: Denisenko and colleagues demonstrated preparation and storage biases in adult mouse kidney [4]. The direction and magnitude cannot be assumed identical in human tumors or other tissues.
A useful pre-collection check is to compare the proposed sampling map with the expected location of the target population. If stromal remodeling is the endpoint, a tumor-rich fragment with little surrounding stroma may be a poor choice. If viable tumor states are the endpoint, extensive acellular necrosis may reduce useful recovery. These are design considerations, not automatic exclusions. Record who selected the region and what evidence informed that choice. After processing, compare recovered cell identities with the expected compartment and available histology. Unexpected absence of a population should prompt review of sampling and preparation before being interpreted as biological absence.
Tissue mass alone cannot predict usable cell yield. Cell density, target-population abundance, preservation, and processing losses all contribute. Where feasible, retain reserve tissue and document a pilot outcome. A photograph and adjacent section provide useful context, but neither fully verifies the exact composition of the processed fragment.
Record collection-to-preservation time, preparation duration, and batch assignment. Balance processing across groups to reduce confounding. A perfect schedule may be impractical, so document deviations and assess their impact. Confirm format-specific transport requirements through sample submission guidelines and frozen tissue handling guidance. Fresh cells, blood, frozen tissue, and fixed material should not share an assumed universal temperature or shipping rule.
Will the Included Analysis Answer My Research Question?
A list of standard analysis steps is not an endpoint-specific plan. Clustering, annotation, marker identification, and visualization may describe the dataset without testing the primary comparison. Specify whether the study asks about expression within a cell type, changes in cell composition, receptor clonotypes, or a regional difference.
For group comparisons, the statistical design must account for independent specimens and cells nested within them. Treating cells as independent donor replicates can inflate false discoveries [2]. Pseudobulk analysis, aggregating counts by cell type and specimen, is a common solution. Suitable donor-aware mixed models or other validated approaches may also be appropriate. The requirement is an adequate model of replication and the endpoint, not one mandatory software package.
Before finalizing the scope, specify sample pairing, relevant covariates, multiple-testing control, and handling of cell types with insufficient representation. Clarify whether automated annotations will be reviewed for the target population. For composition analyses, distinguish recovered cell proportions from true tissue abundance: selective preparation losses may affect the former.
Confirm the files required to revisit decisions: raw reads, count matrices, sample metadata, QC summaries, annotations, and agreed analysis objects. Advanced analyses should have defined inputs and limitations. Somatic hypermutation analysis, for example, requires suitable receptor sequence data, not expression counts alone. Discuss endpoint-specific support through RNA-seq data analysis services.
Expression of ligand–receptor genes suggests potential communication, not physical binding. Pseudotime is an inferred ordering, not observed lineage history [1]. Spatial measurements or flow cytometry can corroborate location and identity; causal mechanism claims may require targeted perturbations and functional readouts. An intervention-associated expression change is not by itself proof of direct target engagement.
A Pre-Submission Decision Checklist
Record the five decisions in the project proposal before committing samples. The purpose is to make assumptions reviewable, including where the evidence may remain incomplete.
Figure 3. Five decisions to document before submission: research goal, workflow, quality review, sampling plan, and analysis scope.
| Question | Decision | Record to Retain |
|---|---|---|
| What does new profiling add? | Define missing evidence and appropriate comparisons. | Primary endpoint, reference-data review, replicate rationale. |
| Which workflow fits? | Match required outputs and available material. | Chemistry, input format, library types, limitations. |
| What happens at quality review? | Agree on pause, optimization, replacement, or risk continuation. | Review criteria, authorization, fees, fallback. |
| Which tissue should be collected? | Define regions and identity retention. | Sampling map, preservation history, batch plan. |
| Will analysis test the endpoint? | Specify donor-aware comparisons and validation needs. | Models, deliverables, scope of revisions. |
For a feasibility review, provide the research question, groups, independent specimen counts, preservation format, collection regions, and target outputs. Explore research-use sequencing services to discuss a project-specific workflow.
Frequently Asked Questions
References
- Heumos L, Schaar AC, Lance C, et al. Best practices for single-cell analysis across modalities. Nature Reviews Genetics. 2023;24(8):550–572.
- Squair JW, Gautier M, Kathe C, et al. Confronting false discoveries in single-cell differential expression. Nature Communications. 2021;12:5692.
- Hsu CY, Liu Q, Shyr Y. A distribution-free and analytic method for power and sample size calculation in single-cell differential expression. Bioinformatics. 2024;40(9):btae540.
- 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 Biology. 2020;21(1):130.
- Salcher S, Heidegger I, Untergasser G, et al. Comparative analysis of 10X Chromium vs. BD Rhapsody whole transcriptome single-cell sequencing technologies in complex human tissues. Heliyon. 2024;10(7):e28358.
- Langerud J, Eilertsen IA, Moosavi SH, et al. Multiregional transcriptomics identifies congruent consensus subtypes with prognostic value beyond tumor heterogeneity of colorectal cancer. Nature Communications. 2024;15:4342.
- BD Biosciences. BD Rhapsody System TCR/BCR Full Length and mRNA Whole Transcriptome Analysis (WTA) Library Preparation Protocol. Manufacturer protocol 23-24017(01), January 2022. Accessed October 12, 2026.
- 10x Genomics. Cell Preparation for Single Cell Protocols. Manufacturer guidance. Accessed October 12, 2026.