‘Capturing the complexity of single-cell interactions’ – Written by Charlotte Harrison, Science Writer.
Single-cell and spatial genomics approaches generate high-dimensional datasets that enable researchers to analyse gene expression. Yet such analysis is typically done one gene at a time, meaning it can miss the complexity and cellular connections that facilitate biological processes.
Now, a team of researchers from the University of California, Irvine have developed an analysis method that captures more of the complexity of how individual cells interact with each other. The method has the, rather complex, name ‘hdWGCNA’ — high-dimensional weighted gene co-expression network analysis — and can be used to investigate systems-level changes from single-cell or spatial transcriptomic datasets. The method is open-source and modular, enabling analysis of cellular or spatial hierarchies in a technology-agnostic manner.
Meta-cell analysis
The authors aimed to overcome two key problems with single-cell and/or transcriptomic data: the inherent sparsity and noise, and the variation in the correlation structure of data from different cell types, cell states and anatomical regions.
The typical hdWGCNA workflow for scRNA-seq data accounts for these problems by collapsing data from cells that are very transcriptomically similar into so-called ‘metacells’. The authors note that this method reduces sparsity while retaining cellular heterogeneity and enables a modular design so that separate network analyses can be performed in specified cell populations.
They note that the hdWGCNA method provides functions for inferring gene networks, identifying gene modules and performing gene enrichment analysis, statistical tests and visualising data. In addition, hdWGCNA can perform isoform-level network analysis using long-read single-cell data.
A delve into Alzheimer’s disease
To demonstrate the utility of the method, the authors performed a systems-level analysis of microglia snRNA-seq data from three human Alzheimer’s disease datasets. Microglia are the immune cells of the brain, and they likely have a role in the pathogenesis of Alzheimer’s disease, where they transition from a homeostatic role to a neuron-damaging role.
Here, the authors unearthed previously unknown gene signatures that were linked to key microglial processes such as axon guidance, phagocytosis, neutrophil activation and myeloid cell differentiation, with CD163 acting as the hub gene.
“Our network analyses of the Alzheimer’s disease datasets show that hdWGCNA is capable of uncovering expanded disease-relevant gene sets,” said the authors.
On a technical level, the hdWGCNA R package directly extends the familiar Seurat pipeline and the SeuratObject data structure; this feature means that other researchers will be able to easily incorporate hdWGCNA network analysis into their own workflows.
