Diagnosis: Why spatial transcriptomics sample results often miss the mark
I remember the evening of April 12, 2021, at our Cambridge bench—eight human liver sections loaded, twelve hours of runs queued, and only three slides cleared QC. The stereo-seq sample gallery had examples that looked perfect (beautiful images, crisp gene maps), yet my runs showed a 40% drop in UMI capture—what broke down between sectioning and sequencing? This scenario + data + question frames the day I started treating workflows like detective work: scenario (core lab run), data (3/8 passed QC), question (where did those transcripts go?).
Over my fifteen-plus years working with spatial omics, I’ve seen the same culprits again and again: incomplete fixation, RNase exposure during transfer, barcode collision, and inadequate sequencing depth — all hurting spot resolution and mapping rates. I’ll be blunt: many “quick fixes” people reach for (longer permeabilization, higher PCR cycles) patch a symptom and worsen others. Once, swapping a commercial fixation buffer for a freshly made 4% paraformaldehyde solution rescued a hippocampus section run; UMI recovery jumped by roughly 30% in that batch. These are the traditional solution flaws I focus on—band-aids that obscure root causes rather than solving them. That insight steers us toward measurement; next, I outline what truly matters.
Forward view: Metrics, comparisons, and real choices (technical)
When I compare fixes, I break things down into concrete metrics: mapping rate, median UMIs per spot, and effective spot resolution. Measuring these lets you separate noise from signal. For instance, increasing sequencing depth improved gene detection in one kidney dataset I handled in 2022 but did not fix low mapping rates caused by barcode misassignment — different fault, different remedy. You should track sequencing depth, barcode integrity (check for collisions), and per-spot UMI distribution; those numbers tell a clearer story than visual inspection alone.
What’s Next — which path to choose?
Here’s how I decide. First, I audit sample prep (date-stamped logs, storage temps, reagent batch numbers). Second, I run a rapid pilot: two slides with modified permeabilization times and one control. Third, I compare three metrics side-by-side and pick the workflow that raises mapping rate without sacrificing UMI balance (short, measurable tests — no guessing). I’ve done this on Stereo-seq chips and on slide arrays; the method holds. Also — keep a simple lab notebook: one misplaced temp reading once cost us a week of reruns; don’t be that team.
To close with practical measures: evaluate solutions by (1) mapping rate (percent reads assigned to spots), (2) median UMIs per spot (signal strength per location), and (3) reproducibility across replicates (consistency). I recommend running these metrics routinely and automating their plotting if you can — it saves time and reveals trends. I’ll admit, I still enjoy the hands-on troubleshooting; it keeps me honest. For more example outputs and reference patterns, check the spatial transcriptomics sample results in the stereo-seq sample gallery — they helped me refine thresholds early on. That said, I expect you’ll adapt thresholds to your tissue and experiment type. A short pause here: try the three metrics above, and then iterate. Finally, if you want a curated starting point for protocols and reference images, I often point colleagues to stomics for examples and tools.