Question 1

How many cells and how many donors does a study need?

Accepted Answer

Cell number governs the detection of rare populations, while donor number governs whether a condition comparison holds. Many studies over-invest in cells and under-invest in biological replicates, leaving comparisons underpowered. Decide the smallest population that must be detected and the effect size across conditions, then size both dimensions against those targets.

Question 2

Should whole cells or nuclei be profiled?

Accepted Answer

Whole cells give more transcripts per barcode and retain cytoplasmic messages. Nuclei suit frozen archival material, fibrous tissue and large or fragile cell types that dissociation destroys, and they carry less stress signature. The two are not directly comparable, so keeping one format across a study avoids confounding the main comparison.

Question 3

How is dissociation-induced stress recognised?

Accepted Answer

Immediate early genes and heat shock transcripts rise sharply during warm enzymatic digestion, producing clusters that reflect handling rather than biology. Cold-active protease protocols and transcription inhibitors reduce the effect. Scoring cells for a stress signature and reporting it openly is preferable to silently removing the clusters it produces.

Question 4

How should clustering resolution be chosen?

Accepted Answer

Resolution is a decision about granularity, not a value to be discovered. Examine stability across a range of settings, check whether candidate clusters are separated by markers with plausible biology, and validate borderline splits with independent evidence. Report the chosen setting and show how conclusions behave when it is varied.

Question 5

What is the correct way to test differential expression between conditions?

Accepted Answer

Treating individual cells as independent replicates inflates significance dramatically, since cells from one donor are correlated. Aggregating counts per donor within each cell type and testing across donors, or fitting mixed models that include a donor term, respects the design. Report the number of donors alongside the number of cells.

Question 6

How much integration across batches is appropriate?

Accepted Answer

Integration methods can remove real biological differences along with technical variation, particularly when conditions and batches are confounded. Design the experiment so each batch contains samples from every condition. When correcting, compare integrated and uncorrected embeddings and confirm that biological signal of interest survives the procedure.

Question 7

What independent evidence supports a newly described population?

Accepted Answer

Marker expression in the same dataset is circular on its own. Stronger support comes from detecting the population in an independent cohort, localising it in tissue by imaging or spatial profiling, isolating it by flow sorting for functional assay, or recovering it in published reference atlases processed independently.

Biotechnology

Biotechnology

Research Frontiers in Single-Cell Transcriptomics Analysis