Design and power analysis for multi-sample single cell genomics experiments

Abstract

Background:The identification of genes associated with specific experimental conditions, genotypes or phenotypes through differential expression analysis has long been the cornerstone of transcriptomic analysis. Single cell RNA-seq is revolutionizing transcriptomics and is enabling interindividual differential gene expression analysis and identification of genetic variants associated with gene expression, so called expression quantitative trait loci at cell-type resolution. Current methods for power analysis and guidance of experimental design either do not account for the specific characteristics of single cell data or are not suitable to model interindividual comparisons. Results: Here we present a statistical framework for experimental design and power analysis of single cell differential gene expression between groups of individuals and expression quantitative trait locus analysis. The model relates sample size, number of cells per individual and sequencing depth to the power of detecting differentially expressed genes within individual cell types. Power analysis is based on data driven priors from literature or pilot experiments across a wide range of application scenarios and single cell RNA-seq platforms. Using these priors we show that, for a fixed budget, the number of cells per individual is the major determinant of power. Conclusion: Our model is general and allows for systematic comparison of alternative experimental designs and can thus be used to guide experimental design to optimize power. For a wide range of applications, shallow sequencing of high numbers of cells per individual leads to higher overall power than deep sequencing of fewer cells. The model is implemented as an R package scPower.

Aging-related olfactory loss is associated with olfactory stem cell transcriptional alterations in humans

“…We performed a prospective analysis of presbyosmic or control normosmic subjects, to obtain nasal mucosal biopsies for analysis. No statistical method was used to predetermine sample size; rather, we have followed current best practices for scRNA-seq studies, and, guided by goals for rare cell type identification, expected read depth, and identification of DE genes for specific clusters, utilized three subjects in each cohort (61). Tissue samples from a single patient were processed individually; as was a single cell culture sample for comparison.…”

Aging-related olfactory loss is associated with olfactory stem cell transcriptional alterations in humans

“…We performed a prospective analysis of presbyosmic or control normosmic subjects, to obtain nasal mucosal biopsies for analysis. No statistical method was used to predetermine sample size; rather, we have followed current best practices for scRNA-seq studies, and, guided by goals for rare cell type identification, expected read depth, and identification of DE genes for specific clusters, utilized three subjects in each cohort (45). Tissue samples from a single patient were processed individually.…”

Applications of single-cell genomics and computational strategies to study common disease and population-level variation

“…For example, based on the user-specified frequency of the rarest cell population and the number of populations with approximately this frequency, SCOPIT (Davis et al 2019) can estimate the number of cells for planning single-cell sequencing experiments. Schmid et al (2020) developed scPower, a more general framework for single-cell power calculation, in which they showed that, for a fixed budget, the number of cells per individual is the major determinant of power of detecting rare cell types and differentially expressed genes, followed by the number of subjects and read depth. In general, shallow sequencing of high numbers of cells per individual leads to a higher overall power than does deep sequencing of fewer cells.…”