Skip to main content
Fig. 1 | Genome Biology

Fig. 1

From: scMET: Bayesian modeling of DNA methylation heterogeneity at single-cell resolution

Fig. 1

Graphical outline for scMET. a Overview of the scMET probabilistic graphical model. The random variables and data that form the model, along with the distributional assumptions, are shown. Input values are denoted by gray circles. Model parameters are denoted by white circles. b scMET uses single-cell DNAm data as input. The data could consist of measurements obtained from different groups of cells, such as experimental conditions or cell types (represented by green and orange colors in the diagram). For each region of interest (e.g., promoters), the input data is recorded in terms of the number of CpG sites for which a valid measurement was recorded and, among those, the number of methylated CpG sites. Note that many CpG sites will not be covered by a read (denoted by red color), leading to sparse information per genomic region. c By combining a hierarchical beta-binomial specification with a generalized linear model framework, scMET captures the mean-overdispersion relationship (left) that is typically observed in bisulphite sequencing readouts and derives residual overdispersion estimates that are not confounded by mean methylation (right). d scMET can be used to identify HVFs that drive epigenetic heterogeneity within a cell population. For example, these could be used as the input of dimensionality reduction techniques or clustering analyses. e scMET uses a probabilistic decision rule to perform differential methylation analysis: to identify features that show differences in mean methylation (left) and/or methylation variability (right) between pre-specified groups of cells

Back to article page