Assessment of dispersion metrics for estimating single-cell transcriptional variability
Tina Chen, Laurie A. Boyer, Divyansh Agarwal
Abstract
Single-cell RNA sequencing data enables analysis of transcript levels of single cells across different cell types and conditions. Recent work has highlighted the value of measuring gene-specific transcriptional variability, or noise, within a genetically identical population of cells in addition to mean expression, given that these differences contribute to biological processes including development and disease.
Introduction
Single-cell RNA sequencing (scRNA-seq) approaches have enabled the study of gene expression in individual cells to better appreciate stochastic biological processes and precisely map cell states. Despite its evolution over the last decade, scRNA-seq data remain sparse in any given experiment, with a large proportion of dropouts, which occur when transcripts are not captured due to efficiency and sequencing depth challenges.
Methods
Generating simulated scRNA-seq data
The sampling distributions for our simulations were chosen for their ability to model single-cell RNA-seq counts. For example, the Poisson(λ) distribution models the probability that k UMIs align at a given locus, assuming UMI counts follow a Poisson process. Distributions like the negative binomial and beta-Poisson can be thought of as a mixture of a measurement model and an observation model, where the Poisson models the reads captured by scRNA-sequencing, and the gamma and beta distributions respectively model the true expression of reads.
Results
Simulations assess relative sensitivity to variability in single-cell counts
We first compared commonly used dispersion metrics for the quantification of transcriptional variability. The Gini index, which is often used to measure economic inequality; the VMR or Fano factor, a measure of deviation from the Poisson distribution; Shannon entropy, a measure of uncertainty in information; coefficient of variation (CV), a measure of standardized dispersion; and squared coefficient of variation (CV2) are metrics that have been used for quantifying transcriptional noise [17,25] and clustering [26,27].
Discussion
Here we performed a comparison of the sensitivity and behavior of six metrics for quantifying dispersion in single-cell data to enable more robust quantification and study of transcriptional variability. We also measured transcriptional variability in three scRNA-seq datasets across two sequencing technologies and demonstrated that it can be used to identify candidate genes and pathways that can broaden our understanding of biological processes.
Acknowledgments
We thank John H. Day in the Boyer lab for helpful discussions.
Citation: Chen T, Boyer LA, Agarwal D (2026) Assessment of dispersion metrics for estimating single-cell transcriptional variability. PLoS Comput Biol 22(3): e1014030. https://doi.org/10.1371/journal.pcbi.1014030
Editor: Zhixiang Lin, The Chinese University of Hong Kong Faculty of Science, HONG KONG
Received: May 24, 2025; Accepted: February 16, 2026; Published: March 2, 2026
Copyright: © 2026 Chen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: UMI counts from Manivannan et al. are available at GEO with accession number GSE193746. UMI counts from Lana-Elola et al. (10.1126/scitranslmed.add6883) are available at GEO with accession number GSE196447. UMI counts from Yuzwa et al. (10.1016/j.celrep.2017.12.017) are available at GEO with accession number GSE107122. The code used to generate the figures in this manuscript are available on Github: https://github.com/lboyerlab/Comparison_of_dispersion_metrics/tree/main.
Funding: This work was supported by the National Heart, Lung, and Blood Institute (R01HL140471 to LAB). The funder had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.
Competing interests: The authors have declared that no competing interests exist.
