switchde: inference of switch-like differential expression along single-cell trajectories

Abstract

Motivation: Pseudotime analyses of single-cell RNA-seq data have become increasingly common. Typically, a latent trajectory corresponding to a biological process of interest-such as differentiation or cell cycle-is discovered. However, relatively little attention has been paid to modelling the differential expression of genes along such trajectories.

Results: We present switchde , a statistical framework and accompanying R package for identifying switch-like differential expression of genes along pseudotemporal trajectories. Our method includes fast model fitting that provides interpretable parameter estimates corresponding to how quickly a gene is up or down regulated as well as where in the trajectory such regulation occurs. It also reports a P -value in favour of rejecting a constant-expression model for switch-like differential expression and optionally models the zero-inflation prevalent in single-cell data.

Availability and implementation: The R package switchde is available through the Bioconductor project at https://bioconductor.org/packages/switchde .

Contact: kieran.campbell@sjc.ox.ac.uk.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Sigmoidal expression across pseudotime. (A) The sigmoid curve as a model of gene expression along single-cell trajectories, parametrized by the average peak expression μ₀, the activation strength k and the activation time t₀. (B) An example using the NDC80 gene from the Trapnell dataset (Trapnell et al. (2014)), which had the lowest P-value of all genes tested. Gene expression measurements are shown as the grey points with the maximum likelihood sigmoid fit denoted by the dark line. The maximum likelihood parameter estimates were ${\mu }_g^{(0)}=2.73,\hspace{0.17em}{k}_g=-8.71$ and $t_g^{(0)}=17.61$. (C) Zero-inflated differential expression for the transcription factor MYOG. Solid line shows the MLE sigmoidal mean while crosses show imputed gene expression measured as zeroes. (D) Posterior predictive density for the zero-inflated model with the solid line denoting MLE sigmoidal mean.