RigorMed
← Research resources

Life and environmental sciences

How to plan pseudobulk analysis for single-cell differential expression

Identify biological replicates before estimating a treatment effect. Aggregate cells within each sample, condition, and cell type, then compare samples using the study design.

Life and environmental sciences

Gather raw counts and cell metadata, including cell ID, sample, donor, condition, batch, and cell type. Build a sample sheet to see whether treatment and batch overlap.

Pseudobulk commonly sums counts within a sample. Save the number of cells contributing to each sample, the filtering rules, and the target cell type.

How to do it

  1. Define the comparison

    Decide whether the outcome is expression within a cell type or the proportion of that type. Expression analysis starts from sample-level counts for the same cell type.

  2. Check biological replication and pairing

    List sample, donor, condition, and batch. Retain a paired donor ID where both conditions were measured. Record technical replicates, libraries, and cells separately from donor counts.

  3. Aggregate and assess eligibility

    Sum raw gene counts for each sample and cell type to form a gene-by-sample matrix. Document low cell counts and absent types. State filtering thresholds and their basis.

  4. Review the fitted comparison against the design

    Use an appropriate count model and normalization. Account for pairing or batch where the design permits. Report included samples, donors, and genes. Check for complete confounding of batch and condition.

Worked example

Two synthetic donors each have control and treated samples. The table shows summed raw counts for one gene in the same cell type.

Worked example
Donor and conditionCellsSummed counts
D1, control50140
D1, treated62202
D2, control44103
D2, treated57149

The matrix retains four sample columns and two paired donors. Cell count and library size may affect raw totals. A gene total alone is insufficient to interpret a treatment effect.

Common questions

Can I combine all cells in an arm into one column?

That discards between-sample variation. Retain one column per independent sample and document donor and condition.

What if a sample has no cells of the target type?

Record absence and filtering reasons first. A missing cell type differs from observed cells with a zero count for a gene.

Verified sources

The official sources below were checked on 2026-11-13. Access conditions and update dates are rechecked before publication.

  1. Seurat — Differential expression testing and pseudobulk

RigorMed workspace

Audit replication and batch structure

Bring authorized de-identified sample metadata and count definitions. Prepare a donor, pairing, batch, cell-type, and analysis eligibility table.