The lowest tissue input you can run on a nuclei-prep platform — and why it matters more than throughput
For any lab working with scarce or irreplaceable samples, the minimum input is the most consequential number on the spec sheet.
- Minimum input gates everything downstream — if a platform can't accept the tissue you actually have, the sequencing chemistry, library prep, and pipeline never apply.
- Fresh/frozen and FFPE are two separate specs — the Singulator runs fresh or frozen tissue at as little as 2 mg; the Singulator 200+ accepts a 50 µm FFPE curl. Keep the two strictly apart.
- 2 mg is roughly an order of magnitude below the ~20 mg most semi-automated alternatives need — the difference between sampling every donor region and rationing the precious ones.
- Low input doesn't cost yield — ~60,000 nuclei/mg on frozen mouse cortex (Kersey et al. 2026), with consistency that comes from removing operator variability.
"Our cohort is precious. We don't have endless tissue. The PI brought us one block, and we have to make it work."
If you run a single-cell core or a translational program, you've heard a version of this from every PI who walks in with frozen brain, a tumor biopsy, or an archival FFPE block linked to clinical outcomes.
Throughput specs get most of the attention on nuclei-prep platform comparisons. They're worth knowing — and on per-day capacity, the Singulator is competitive with the semi-automated workflows most labs are choosing between (16 samples/hour, ~96/day on a single instrument, non-FFPE). But there's a more fundamental constraint that gates everything else: how little tissue can the platform run and still produce usable data?
If a platform can't accept the tissue you actually have, nothing downstream applies. Not the sequencing chemistry, not the library prep, not the bioinformatics pipeline. The minimum input is the most consequential number on the spec sheet for any lab working with scarce or irreplaceable samples.
What "lowest input" actually means in practice
The answer breaks into two different stories — fresh/frozen tissue, and FFPE — and they should never be mixed up.
Fresh and frozen tissue: The Singulator Platform runs at as little as 2 mg of tissue. That's about an order of magnitude below what most semi-automated alternatives need (~20 mg).
For a BICAN-scale neuroscience cohort, that input differential is the difference between sampling every brain region of every donor and rationing the precious ones. For a translational oncology lab working from a fresh biopsy, it's the difference between having enough for both single-cell AND spatial AND a backup section.
FFPE archives: The Singulator 200+ accepts a 50 µm curl. The diagnostic block stays intact for pathology. One block, sectioned, can feed spatial AND single-cell workflows on the same day.
The 2 mg claim is non-FFPE only. The curl-input claim is FFPE only. They're different specs because they describe different workflows. Keeping that distinction clean matters when you're scoping whether a platform fits your cohort.
How much tissue is actually enough?
When a PI hands you a precious cohort, the calculation isn't "how fast can we run this." It's:
- How many regions of this single donor can I sample?
- How many donors can I afford to lose to a failed prep?
- What's the minimum I can get away with and still have publishable n?
A platform that runs at 2 mg lets a PI sample more regions per donor — or accept smaller donor inputs and still produce data. For consortium-scale work (BICAN, PsychENCODE, HCA), where every donor is a multi-institution coordination effort, the access constraint is an institutional one.
For FFPE archives — where every block was banked years ago for a different purpose, the diagnostic team needs material for downstream pathology, and there is no "go back and get more tissue" option — the curl-input spec is what makes snRNA-seq feasible at all.
What the data shows
For non-FFPE inputs, the Kersey et al. 2026 head-to-head in Cell Reports Methods reported ~60,000 nuclei/mg on frozen mouse cortex — yield parity at the order-of-magnitude lower input.
For FFPE, the PCS PDAC FFPE Application Note (December 2025) reported replicate yields of 1.0M / 1.0M nuclei from 50 µm FFPE curl inputs on mouse PDAC tissue. The manual alternative — a manual column/dissociator-based workflow on the same blocks — produced 1.5M / 0.4M replicates. The Singulator's consistency comes from the fact that the workflow is fully automated end-to-end; output is consistent regardless of input variability, because operator-introduced variability is removed from the equation.
Kersey 2026 is mouse cortex; human-postmortem translation should be caveated until a human-tissue head-to-head is published. What translates: the mechanism — software-controlled mechanical lysis without enzymatic 37°C step — which is independent of species.
When input becomes the gating constraint
For a PI who says "I have one block" or "I have one donor's worth of cortex," the question stops being which workflow is best in head-to-head benchmarks and becomes which workflow lets me do the experiment at all. The lowest input is the gating constraint. Reproducibility, throughput, downstream compatibility — none of it applies if you can't run the tissue you have.
A platform optimized for 20 mg fresh inputs leaves precious-sample cohorts on the table. A platform that runs at 2 mg fresh-or-frozen and a 50 µm FFPE curl doesn't.
Working with a precious cohort? Let's run through your numbers.
20 minutes with a PCS application scientist — walk through your tissue type, your minimum input, your expected donor count. We'll show you exactly how the platform handles your cohort before you commit to anything.
