Maximizing statistical power to detect clinically associated cell states with scPOST

Abstract

As advances in single-cell technologies enable the unbiased assay of thousands of cells simultaneously, human disease studies are able to identify clinically associated cell states using case-control study designs. These studies require precious clinical samples and costly technologies; therefore, it is critical to employ study design principles that maximize power to detect cell state frequency shifts between conditions, such as disease versus healthy. Here, we present single-cell Power Simulation Tool (scPOST), a method that enables users to estimate power under different study designs. To approximate the specific experimental and clinical scenarios being investigated, scPOST takes prototype (public or pilot) single-cell data as input and generates large numbers of single-cell datasets in silico. We use scPOST to perform power analyses on three independent single-cell datasets that span diverse experimental conditions: a batch-corrected 21-sample rheumatoid arthritis dataset (5,265 cells) from synovial tissue, a 259-sample tuberculosis progression dataset (496,517 memory T cells) from peripheral blood mononuclear cells (PBMCs), and a 30-sample ulcerative colitis dataset (235,229 cells) from intestinal biopsies. Over thousands of simulations, we consistently observe that power to detect frequency shifts in cell states is maximized by larger numbers of independent clinical samples, reduced batch effects, and smaller variation in a cell state’s frequency across samples.

Co-varying neighborhood analysis identifies cell populations associated with phenotypes of interest from single-cell transcriptomics

“…To assess CNA’s power and signal recovery, we simulated sample attributes with true associations to different types of cell populations and compared CNA’s performance to that of a cluster-based association test using Mixed-effects modeling of Associations of Single Cells (MASC)[ 13 ]; MASC offers greater power than a t-test or linear model by accounting for per-cell information[ 14 ]. For CNA, power was defined as the proportion of simulations with global p<0.05.…”

Axes of inter-sample variability among transcriptional neighborhoods reveal disease-associated cell states in single-cell data

“…To assess CNA's power and accuracy, we simulated sample attributes with true associations to different types of cell populations and compared CNA's performance to that of a cluster-based association test using Mixed-effects modeling of Associations of Single Cells (MASC) [9]; MASC offers greater power than a t-test or linear model by accounting for per-cell information [10]. For CNA, power was defined as the proportion of simulations with global p<0.05.…”

Designing Single Cell RNA-Sequencing Experiments for Learning Latent Representations

“…Generally, many approaches for designing scRNA-seq experiments rely on simulated data, where researchers either assume linear relationships between genes [20], use kinetic models [21], simple generative models [22], or unsupervised techniques [23], which require large reference data sets.…”

“…Generally, many approaches for designing scRNA-seq experiments rely on simulated data, where researchers either assume linear relationships between genes [20], use kinetic models [21], simple generative models [22], or unsupervised techniques [23], which require large reference data sets. However, sample size estimation tools for learning low-dimensional representations, especially for capturing non-linear relationships, stay mostly unaddressed.…”