A working spatial transcriptomics workflow has more in common with a production line than with a bench experiment. Samples arrive with variable quality, reagents are expensive enough that failures matter, runs occupy instruments for days, and the output needs handing to someone who can analyze it. This section covers the operational layer that determines whether a well-specified platform actually delivers.
Key Takeaways
|
The End-to-End Spatial Workflow
Map the whole chain before optimizing any part of it. Most spatial programs are limited by one stage, and it is rarely the instrument. Ownership matters as much as sequence, because handoffs between groups are where samples and information get lost.
Stage | Typically Owned By | Quality Gate | Common Failure |
Sample intake and consent check | Lab manager or coordinator | Documentation complete, ethics coverage confirmed | Samples accepted that cannot legally be used as intended |
Block assessment and QC | Histology or the spatial team | Quality metric measured and recorded against threshold | Skipped entirely, so failures surface after reagents are spent |
Sectioning and mounting | Histology | Thickness, orientation, and placement within the capture area | Sections cut to a thickness the platform does not accept |
Staining and assay execution | Spatial team | Controls included and reviewed before the run proceeds | No control tissue, so a failed run cannot be diagnosed |
Instrument run | Trained operator | Run parameters recorded, outcome logged | Outcome not recorded, making repeat rate unmeasurable |
Data transfer and archive | Spatial team and IT | Files verified complete and reached managed storage | Data left on the instrument or a local drive |
Analysis handoff | Analyst | Agreed format, metadata, and acceptance criteria | Data delivered that the analyst cannot work with |
Table 1. The end-to-end workflow with its gates and characteristic failures. The second and fifth rows are the two most often missing in practice.
The analysis handoff at the bottom deserves treating as a formal interface rather than an afterthought. Agree in advance what format is delivered, what metadata accompanies it, and what constitutes acceptable data, because an analyst discovering problems weeks later cannot recover the sample. Storage, compute, and the infrastructure behind that handoff are covered in Managing Spatial Biology Data: Storage, Compute, and Infrastructure for Spatial Datasets.
What Should Your Sample Intake Gates Check?
Intake is where the cheapest interventions live. A block screened out at intake costs you a measurement. The same block discovered to be unusable after staining costs you the reagents, the instrument time, and a section of irreplaceable tissue.
The evidence for screening is unusually clear. The technical comparison of spatial transcriptomics platforms across six cancer types published in Genome Biology found a strong correlation between DV200, the proportion of RNA fragments longer than 200 nucleotides, and the average number of genes detected per cell, across platforms. That group screened blocks above 34% DV200 before proceeding. Separately, the systematic benchmarking of imaging platforms in FFPE tissues in Nature Communications notes that platform guidance differs: some vendors suggest pre-screening on stained sections, while MERSCOPE recommends DV200 above 60%.
Gate | What to Check | Action If It Fails |
Provenance and permissions | Consent scope, ethics approval, any restriction on off-site transfer | Stop. This is not a technical problem and cannot be resolved downstream |
Fixation and storage history | Fixative, fixation duration, block age, storage conditions | Flag as elevated risk and adjust expectations rather than rejecting outright |
RNA integrity | DV200 or the metric your platform specifies, against its stated threshold | Reject, or proceed with documented expectations and a larger section allocation |
Morphology | Stained section reviewed for necrosis, folding, and target tissue presence | Re-cut, re-orient, or reject before committing reagents |
Physical fit | Region of interest fits inside the platform capture or imageable area | Trim, re-orient, or plan multiple runs and re-cost the project |
Material sufficiency | Enough tissue remains to absorb at least one repeat | Escalate. A single-section project has no recovery path |
Table 2. Intake gates. The RNA integrity threshold is platform-specific; take it from your vendor rather than adopting a general figure.
Record every gate result whether it passes or fails. Intake records are what let you tell a principal investigator later why their samples underperformed, and they are the only route to an evidence-based conversation about tissue quality rather than a defensive one. Sample handling decisions in detail are covered in FFPE vs. Fresh Frozen for Spatial Biology: Sample Handling Decisions, and QC in depth in Tissue Quality and QC for Spatial Assays: Managing Failure Rates.
Batch Scheduling and Throughput
Scheduling spatial work breaks the habits most managers bring from plate-based assays, because run length is governed by imaged area and panel size rather than by how many samples are queued. Adding samples to a batch may cost nothing. Enlarging the panel or the scanned region can multiply occupancy.
The Genome Biology work provides a concrete illustration: a 377-gene imaging run took roughly 50 hours, while a 5,000-gene run on the same platform and the same samples took about 7 days. That is a thirteenfold panel increase producing roughly a threefold increase in instrument time. Comparable figures for other platforms are collected in Spatial Transcriptomics Platforms Compared: Visium, Xenium, CosMx, MERSCOPE, and GeoMx and Spatial Proteomics Platforms Compared: PhenoCycler, Imaging Mass Cytometry, and MIBI.
Four scheduling rules follow from that.
- Schedule in area and panel size, not sample count. Build your booking system around expected run hours derived from configuration, and require requesters to specify both.
- Batch by configuration, not by requester. Grouping samples that share a panel and section size reduces changeover and makes run length predictable.
- Protect a repeat slot. If your repeat rate is fifteen percent, roughly one run in seven is a rerun. Scheduling at full capacity guarantees that reruns displace new work.
- Do not let a long run start before a weekend shutdown. Multi-day acquisitions cross building maintenance windows and overnight environmental setbacks, both of which can end a run.
Utilization is also the metric that carries most weight in future funding conversations, so instrument the booking system to capture it from the first run rather than reconstructing it later. Where that sits in the wider measurement framework is covered in the Section 1 material on return, and the practicalities of scaling from a pilot are in Scaling Spatial Biology Throughput: From Pilot to Production.
Lab Management Certificate
The Lab Management certificate is more than training—it’s a professional advantage.
Gain critical skills and IACET-approved CEUs that make a measurable difference.
Which Failure Modes Should You Expect?
Plan for failure at a rate rather than hoping for its absence. Published benchmarking gives a realistic picture of what goes wrong, and most of it is upstream of the instrument.
- Tissue loss during processing. In the Genome Biology study, a breast sample partially detached during the 95-degree decrosslinking step, folded, and produced a failed library, which the authors attributed to high lipid content. Fatty and fragile tissues carry real handling risk in sequencing-based workflows.
- Low RNA integrity. Detection tracked sample quality on every platform tested, so a poor block underperforms rather than failing cleanly, which is harder to diagnose.
- Sectioning and placement errors. Wrong thickness, poor orientation, or a region of interest falling outside the capture area. All are recoverable by re-cutting if caught before staining.
- Insufficient material. A project with no spare tissue has no recovery path, which converts an ordinary failure into a lost study.
- Environmental interruption. Power or temperature excursions during a multi-day run, which is why long runs need protected conditions.
Two practices reduce the cost of all of these. Include control tissue in every run, because a failed run without controls cannot be diagnosed and you will not know whether to blame the sample, the reagents, or the instrument. And record the cause of every repeat, not merely that a repeat occurred, so the pattern becomes visible before it becomes expensive. Since consumables dominate per-section cost, prevented failures translate directly into budget, as the modeling in How Much Does Spatial Biology Cost? Budgeting for Instruments, Consumables, and Data shows.
Artifacts That Look Like Biology
This is the part of spatial operations that most rewards attention and receives the least. Published benchmarking has documented several platform-specific artifacts that produce plausible but wrong biological conclusions, and none of them announces itself. A lab that knows what to look for has a real QC capability. A lab that does not will publish artifacts.
Artifact | Where It Occurs | How It Presents | What to Do |
Bin-level cell mixing | Sequencing-based, at aggregated bin sizes | Transcripts from 2 to 4 cells at 8 micron bins, and 5 to 10 cells at 16 micron bins | State the bin size used in every result, and interpret cell-level claims cautiously |
Composition bias by cell size | Sequencing-based binned data | Epithelial and stromal cells over-represented relative to immune cells, because larger cells span more bins | Do not read binned data as a cell census without correction |
Grid-aligned striping | Sequencing-based high-definition data | Regular intensity variation aligned with the underlying barcode grid | Inspect images for periodic patterns before interpreting spatial structure |
Field-of-view edge depletion | Imaging platforms using discrete fields of view | Transcript counts near a field edge an order of magnitude lower than the interior | Exclude an edge margin, or verify the vendor pipeline already does |
Background and negative probe signal | Imaging platforms, varying widely between them | Low-abundance transcripts indistinguishable from background where negative probe rates are high | Check the negative probe rate on your own data and treat low-abundance calls sceptically |
Segmentation error | Imaging platforms | Transcripts assigned to the wrong cell, inflating apparent cell size in sparse regions | Review segmentation visually, and prefer membrane-informed segmentation where available |
Small-region sampling bias | Any platform, when few small regions are imaged | Composition estimates ranging from 5% to 55% against a whole-section value of 22% | Image adequate area, and never report composition from a handful of small fields |
Table 3. Documented artifacts from the two benchmarking studies cited in this article. Figures reflect the platform configurations those studies tested, mid-2025 for the Genome Biology work.
The Case for Making This a Standing QC Step The last row is the most consequential and the least technical. In the Genome Biology study, a single 0.5 by 0.5 millimetre field of view produced cancer cell proportion estimates anywhere from 5% to 55%, while the whole section value was 22%. Any composition claim derived from a few small regions is unreliable, and no amount of downstream statistics fixes it. Build a short visual review into the workflow after every run: check for periodic patterns, inspect field edges, review segmentation on a sample of cells, and confirm the imaged area is adequate for the claims intended. Ten minutes of looking prevents months of misinterpretation, and it costs nothing but attention. |
Interested in life sciences?
What Should Your SOPs Actually Document?
SOPs for spatial work fail in two directions. Too thin, and nothing is reproducible when the person who developed the method leaves. Too heavy, and nobody follows them. Aim for a small set of documents that each answer a question someone will actually ask.
Document | What It Must Specify |
Sample acceptance criteria | Quality thresholds, required documentation, and who has authority to accept a marginal sample |
Sectioning specification | Thickness, orientation, mounting, placement within the capture area, and storage of cut sections |
Assay protocol | The run procedure, controls required in every run, and stop and go decision points |
Run log standard | What is recorded for every run: configuration, operator, outcome, and cause of any repeat |
Data handling | Where files land, how completeness is verified, retention period, and who owns the archive |
Analysis handoff | Delivery format, accompanying metadata, and acceptance criteria agreed with the analyst |
Artifact QC checklist | The visual review performed after every run, and what triggers escalation |
Training and competency | What a new operator must demonstrate before running samples unsupervised |
Table 4. A minimum viable SOP set. Each document exists to answer a recurring question rather than to satisfy an audit.
Two design principles keep them usable. Write them for the person who will do the work rather than for a reviewer, which usually means shorter and more specific than instinct suggests. And version them, because a protocol that changes without a record makes historical data uninterpretable. What to document in full detail is covered in Standard Operating Procedures for Spatial Assays: What to Document, and the scheduling, chargeback, and access questions specific to shared facilities are in Managing a Spatial Biology Core Facility: Scheduling, Chargebacks, and Access.
Underpinning all of it is capacity. Workflow discipline cannot substitute for staff who have allocated time or for infrastructure that can absorb the data, and the four-domain view of whether those exist is in Is Your Lab Ready for Spatial Biology? A Readiness Assessment. For the operational picture across procurement, workflow, data, and staffing together, see Spatial Biology in the Lab: A Manager's Guide to Evaluating, Implementing, and Scaling Spatial Technologies.
This article was produced under Lab Manager's AI Editorial Guidelines.
