The smallest sample in your core: standardized single-cell prep down to 2 mg
When the sample is a needle biopsy, a sorted population, or a few milligrams of rare tissue, there is no second curl if the prep fails. Here is what it takes to turn inputs that small into sequencing-ready cells or nuclei without grinding the material away, and how to make it a standard your whole core can run.
- The Singulator is validated for inputs as low as 2 mg — the scale of needle biopsies, sorted populations, and rare tissue a shared core cannot afford to waste on a failed prep. On the Singulator 200+, a 50 µm FFPE curl runs through to nuclei on the same platform.
- Low input is where gentle chemistry matters most. Cold, enzyme-free processing and short run times protect scarce material instead of shearing it, and the same software-controlled protocol runs whoever is at the bench.
- Reproducibility is the proof that matters here. In peer-reviewed testing the platform gave the lowest sample-to-sample variability of the methods compared, so the result holds up across operators and across the small samples you only get once.
The smallest sample is the hardest to prep, because every loss is proportional but the consequences are not. Lose 30% of a 300 mg block and you still have plenty of cells. Lose 30% of a 2 mg needle biopsy and you may fall below the input a downstream platform needs for a usable library, with no replicate to average and no spare to retry.
A single-cell genomics core does not get to choose its intake. Some weeks the precious sample is a core-needle biopsy from a research study, a FACS-sorted subset that took a day to collect, or a slice of rare patient tissue with no replacement in the freezer. The PI has one shot.
Two things make small samples fragile in ways a generous sample hides. Every tube-to-tube transfer, wash, and filtration step strands a fraction of the cells or nuclei on plastic, and on a few milligrams that fraction is the difference between a run and a write-off. And operator variance has nowhere to hide: across many large samples a senior scientist's technique and a new hire's average out, but on a one-of-one biopsy, whoever is at the bench that day sets the entire result. So an ultra-low-input prep has to do three things: minimize hands-on transfers, handle the material gently enough that recovered cells still represent the tissue, and run identically regardless of who loads it. The question is whether that can be done on the same instrument a core already runs for everything else.
In brain, enzymatic dissociation at 37 °C can induce artifactual stress-response and microglial-activation signatures (Marsh et al., 2022, Nature Neuroscience). That is a property of warm enzymatic handling, not of any one instrument. It is also why a cold, enzyme-free nuclei workflow starts from a better place: less artifact to introduce, and less material lost to a harsh digest you cannot afford on a small sample.
Marsh et al. is cited here as context for why gentle, cold prep matters, not as a Singulator result.
What the Singulator does at 2 mg
The Singulator is an automated single-cell sample-prep system that turns fresh, frozen, OCT, and FFPE tissue into sequencing-ready cells or nuclei on a software-controlled instrument. For the low-input case, three design choices carry the weight.
Validated inputs as low as 2 mg, and a 50 µm FFPE curl
The platform is validated for fresh or frozen inputs as low as 2 mg, which covers the samples a core most often hesitates to accept. Separately, on the Singulator 200+, a 50 µm FFPE curl runs through to nuclei, so the archival block where only a thin curl can be spared has a path on the same instrument. The two are different specs for different workflows: 2 mg is the fresh-and-frozen low-input figure, and the 50 µm curl is the FFPE path on the 200+. Keeping them apart is what keeps each one accurate.
Cold, enzyme-free chemistry in an enclosed cartridge
The nuclei side runs cold and enzyme-free, with integrated 4 °C control, inside an enclosed single-use cartridge with inline filtration and no internal fluidics. A scarce sample spends less time being handled, and the enclosed cartridge removes a class of loss and contamination that open, multi-tube manual workflows introduce.
The same protocol, whoever is at the bench
Timing, force, temperature, and chemistry are set in software and run identically every time. On a one-of-one sample, where there is no margin to absorb a difference in technique, that matters more than anywhere else. The validated method lives in the instrument and a native run log records every run, so the prep that succeeded on a precious sample is documented and repeatable for the next one.
Consistency is the proof for a small sample
The evidence for a low-input claim is consistency: the same quality of result every time, including on the samples you only get once. A single big yield figure tells you far less.
In a head-to-head comparison of nuclei isolation methods on frozen mouse cortex, the Singulator gave nuclei that were roughly 100% structurally intact, versus about 85% for a sucrose gradient and about 35% for a column kit. The same nuclei carried under 0.5% mitochondrial reads, among the lowest ribosomal contamination of the methods compared, and the lowest sample-to-sample variability of any method tested (Kersey et al., 2026, Cell Reports Methods).
Reproducibility holds at the level of yield too. In a mouse PDAC study, technical replicates returned identical yield (1.0M and 1.0M nuclei) where a manual workflow swung close to four-fold between replicates (1.5M and 0.4M); n = 4 from a single mouse PDAC block (PCS FFPE application note, 2025).
Lowest sample-to-sample variability is the claim that matters for a small sample: it means the result depends on the sample, not on the run.
On a 2 mg sample, low variability is the property that protects you. The platform preserves the cell-type and cell-state representation of the input rather than skewing it. That is what lets a single precious sample stand in for the biology the PI set out to study.
Low input doesn't have to mean low yield
Will ultra-low input mean low yield or compromised quality?
Low input is exactly where gentle chemistry pays off: cold, enzyme-free processing and validated inputs to 2 mg protect usable material rather than grind it away, and the peer-reviewed variability and integrity data hold on small samples. For any new low-input tissue, run a quick QC check against your own baseline on the first samples, with support from PCS applications scientists, so you confirm the result on your material before you commit a precious one.
Does the same instrument handle a needle biopsy and a 300 mg block?
Yes. The platform covers low-input samples validated to 2 mg, up to several hundred milligrams across its cartridge menu, so a core runs the biopsy and the generous block on one footprint without standing up a separate low-input method.
Can it prep a sorted or otherwise pre-purified population?
Yes. Sorted subsets are a common low-input case, handled with the same gentle, enclosed, low-transfer processing, so the population you spent a day collecting is not lost to handling between the sorter and the sequencer. The native run log documents the prep for provenance, which matters when the input is irreplaceable.
How do I trust a result on a sample I cannot repeat?
Validate the method on routine material first, where you can run replicates against your baseline. Then apply the exact same software-controlled, logged protocol to the precious sample, backed by reproducibility documented in peer-reviewed data, so you are not gambling the irreplaceable sample on a first attempt.
Where this already runs
Standardized prep on one footprint is how multi-user cores already operate. The Yale Center for Genome Analysis (YCGA) runs scRNA-seq, snRNA-seq, ATAC-seq, CITE-seq, and FACS off a single Singulator footprint, across the range of samples its investigators bring in. Validated across a broad range of tissue types, the platform is built for a core that takes in whatever the next PI submits, including the smallest and most precious of it. The same standardization argument runs underneath the case for treating prep as the most variable step and the case for one platform across every sample type; if the worry is whether one instrument can hold the standard across every operator, that objection is answered here.
Your intake queue changes every day. Your prep quality shouldn't, and that holds most on the sample you only get once.
