Impulse model-based differential expression analysis of time course sequencing data

Abstract

Temporal changes to the concentration of molecular species such as mRNA, which take place in response to various environmental cues, can often be modeled as simple continuous functions such as a single pulse (impulse) model. The simplicity of such functional representations can provide an improved performance on fundamental tasks such as noise reduction, imputation and differential expression analysis. However, temporal gene expression profiles are often studied with models that treat time as a categorical variable, neglecting the dependence between time points. Here, we present ImpulseDE2, a framework for differential expression analysis that combines the power of the impulse model as a continuous representation of temporal responses along with a noise model tailored specifically to sequencing data. We compare the simple categorical models to ImpulseDE2 and to other continuous models based on natural cubic splines and demonstrate the utility of the continuous approach for studying differential expression in time course sequencing experiments. A unique feature of ImpulseDE2 is the ability to distinguish permanently from transiently up- or down-regulated genes. Using an in vitro differentiation dataset, we demonstrate that this gene classification scheme can be used to highlight distinct transcriptional programs that are associated with different phases of the differentiation process.

Current best practices in single‐cell RNA‐seq analysis: a tutorial

“…The ordering of cells along these paths is described by a pseudotime variable. While this variable is related to transcriptional distances from a root cell, it is often interpreted as a proxy for developmental time (Moignard et al, 2015;Haghverdi et al, 2016;Fischer et al, 2018;Griffiths et al, 2018). Since Monocle (Trapnell et al, 2014) and Wanderlust (Bendall et al, 2014) established the TI field, the number of available methods has exploded.…”

Identification of transcription factors regulating senescence in wheat through gene regulatory network modelling

“…In total 52,905 genes met 515 this condition. We used the count expression level of these genes for differential expression analysis 516 using the R package ImpulseDE2 v1.4.0 (Fischer et al, 2018), all counts were rounded to the nearest 517 integer before they were analysed with ImpulseDE2. In parallel we used the TPM expression level of 518 these 52,905 genes for differential expression analysis using Gradient Tool v1.0 (Breeze et al, 2011) 519 with the normalisation enabled on Cyverse (https://de.cyverse.org/de/) (Merchant et al, 2016).…”

Dynamic metabolomic analysis of intestinal ischemia–reperfusion injury in rats

“…The principal component analysis (PCA), a dimension reduction method, was applied to explore grouping trends and outliers in datasets. Different features were identified by ImpulseDE2 (https://github.com/YosefLab/ ImpulseDE2) 17 during the time course. Mfuzz soft clustering (https://bioconductor.org/packages/release/bioc/ html/Mfuzz.html) 18 was used to uncover major temporal clusters of the significantly changed molecules.…”