Optimising Sample Size and Cell Number for Single Cell Analysis

26.02.2025

2’

When setting up a single cell analysis project, many researchers focus on sample numbers per group. However, statistical power—the ability to detect true biological differences—often gets overlooked.

In single cell analysis, we measure thousands of cells in each sample. But how many samples should we include per group, and how many cells of interest do we need per sample? These questions can be answered by looking into statistical power of your experiment.

Why Statistical Power Matters

Statistical power helps ensure that you can detect meaningful changes in gene expression when they exist. Factors that increase power include:

Number of samples per group

Number of cells from the subpopulation of interest

Magnitude of gene expression differences

If you have too few samples or not enough cells, you may fail to detect important differences. This is especially true for less abundant sub-populations or subtle changes in gene expression.

Key Finding from Our Simulations

We ran simulations using bone marrow mononuclear cell data to see how cell numbers and sample replicates affect statistical power. We found that to reliably detect a three-fold difference in gene expression, it usually takes > 500 cells per sample per population of interest if you have three to four samples per group i.

That translates to a few thousand cells per sample for abundant cell types, or 10,000 cells and more for cell populations accounting for <5% of the total.

If you are working with fewer cells per sample or suspect more subtle expression differences, you may need more samples or higher cell numbers to achieve acceptable power (commonly 0.8 or higher).

Putting It into Practice

Estimate Your Subpopulation: Determine how many cells of your target population you can capture in each sample.

Set Target Replicates: Based on your expected effect size, plan the number of samples per group.

Pilot If Necessary: A pilot run can help confirm cell yields and guide final decisions.

Get the Flyer

For more detail on how we ran our simulations—and how you can apply these insights to your own project—check out our Statistical Power Analysis Flyer. It’s a short read that might save you time and resources later.

Download here