# A distribution-free and analytic method for power and sample size calculation in single-cell differential expression

## TL;DR

This paper introduces scPS, a new method for calculating statistical power and sample size in single-cell RNA sequencing experiments without assuming data distribution.

## Contribution

The novel contribution is an analytic method for power and sample size calculation that does not assume data distribution and accounts for cell-cell correlations.

## Key findings

- scPS provides accurate power and sample size estimates without assuming data distribution.
- The method considers cell-cell correlations within individual samples, improving biological accuracy.
- scPS is computationally efficient compared to simulation-based approaches.

## Abstract

Differential expression analysis in single-cell transcriptomics unveils cell type-specific responses to various treatments or biological conditions. To ensure the robustness and reliability of the analysis, it is essential to have a solid experimental design with ample statistical power and sample size. However, existing methods for power and sample size calculation often assume a specific distribution for single-cell transcriptomics data, potentially deviating from the true data distribution. Moreover, they commonly overlook cell–cell correlations within individual samples, posing challenges in accurately representing biological phenomena. Additionally, due to the complexity of deriving an analytic formula, most methods employ time-consuming simulation-based strategies.

We propose an analytic-based method named scPS for calculating power and sample sizes based on generalized estimating equations. scPS stands out by making no assumptions about the data distribution and considering cell–cell correlations within individual samples. scPS is a rapid and powerful approach for designing experiments in single-cell differential expression analysis.

scPS is freely available at https://github.com/cyhsuTN/scPS and Zenodo https://zenodo.org/records/13375996.