How long does single-cell analysis take, realistically?
Timelines for each stage of a single-cell project, what actually causes delays, and how to compress a deadline without cutting corners.
Do this on your own data: turnaround depends on which modules are in scope, and that is exactly what a quote pins down. The cost calculator shows scope and price together for your own sample count: omics.hstgenomics.com/cost-calculator. The deliverable those days are spent producing: omics.hstgenomics.com/sample.
Want it done rather than explained? We do analysis only — send a count matrix, an h5ad or Seurat object, Cell Ranger output or a public accession and get publication-ready results with the code that produced them, at a fixed price agreed before any work starts: omics.hstgenomics.com/quote. The same numbers are available to software as JSON, no key and no sign-up: omics.hstgenomics.com/api/quote.
"Two weeks" is the usual answer and it is usually wrong, because most of the elapsed time is not computation. Here is where the days actually go.
Stage by stage
- Data transfer and intake — hours to days. Count matrices move in minutes. Raw FASTQ for a ten-sample project can take a day or more, and institutional transfer restrictions add another. This is the single most compressible delay: send matrices when the question does not require re-alignment.
- Alignment / counting (only if needed) — 1–2 days. Mostly machine time, parallelisable.
- QC, doublets, ambient RNA — half a day. Fast, but this is where problems surface that change the plan.
- Integration, clustering, annotation — 1–3 days. Annotation is the part that needs a person who knows the tissue; it is not a button.
- Differential expression and enrichment — 1 day once the design is agreed.
- Advanced modules (trajectory, communication, spatial deconvolution) — 1–2 days each.
- Figures, tables, methods text — 1 day. Regularly underestimated; making a figure submission-ready takes longer than making it correct.
A standard four-to-eight sample project therefore takes roughly five to eight working days of actual work. Calendar time is longer because of the next section.
What actually causes delays
- Undefined comparisons. "Find what is different" cannot be scheduled. Written contrasts can.
- Missing metadata. Which sample is which subject, which batch, which timepoint. Reconstructing this from file names costs days and introduces errors.
- Discovering a bad sample late. A sample that fails QC changes the design. Finding it on day one is a nuisance; finding it on day six is a restart.
- Review cycles. The analyst sends results, you look at them a week later. This is normally the largest single block of calendar time in the whole project.
How to compress a real deadline
Send count matrices instead of FASTQ. Send the sample table with subject IDs before the data. Write the three comparisons that matter. Agree that the first delivery is figures and tables, with the methods text following a day later. Book your own review slot in advance — a same-day response to a first-pass result routinely saves a week.
With all of that in place, a standard single-cell project can go from data to submission-ready figures in about 48 hours of working time. Without it, the same project takes three weeks, and none of the difference is computation.