The ME matrices are stored as a matrix where each row is a cell and each column is a
module. This matrix can be extracted from the Seurat object using the `GetMEs`
function, which retrieves the **hMEs by default**.
```{r eval=FALSE}
# harmonized module eigengenes:
hMEs <- GetMEs(seurat_obj)
# module eigengenes:
MEs <- GetMEs(seurat_obj, harmonized=FALSE)
```
## Compute module connectivity
In co-expression network analysis, we often want to focus on the "hub genes", those
which are highly connected within each module. Therefore we wish to determine the
*intramodular connectivity*, also known as *kME*, of each gene. scWGCNA includes
the `ModuleConnectivity` to compute the *kME* values in the full single-cell dataset,
rather than the metacell dataset.
```{r eval=FALSE}
# compute intramodular connectivity:
seurat_obj <- ModuleConnectivity(seurat_obj)
```
## Getting the module assignment table
scWGCNA allows for easy access of the module assignment table using the `GetModules`
function. This table consists of three columns: `gene_name` stores the gene's symbol or ID, `module` stores the gene's module assignment, and `color` stores a color mapping
for each module, which is used in many downstream plotting steps. If `ModuleConnectivity` has been called on this scWGCNA experiment, this table will
have additional columns for the *kME* of each module.
```{r eval=FALSE}
# get the module assignment table:
modules <- GetModules(seurat_obj)
```
This wraps up the critical analysis steps for scWGCNA, so remember to save your
output.
```{r eval=FALSE}
saveRDS(seurat_obj, file='scWGCNA_object.rds')
```
### Optional: compute hub gene signature scores
Gene scoring analysis is a popular method in single-cell transcriptomics for computing
a score for the overall signature of a set of genes. Seurat implements their own
gene scoring technique using the `AddModuleScore` function, but there are also
alternative approaches such as [UCell](https://github.com/carmonalab/UCell).
scWGCNA includes the function `ModuleExprScore` to compute gene scores for
a give number of genes for each module, using either the Seurat or UCell algorithm.
```{r eval=FALSE}
# compute gene scoring for the top 25 hub genes by kME for each module
# with Seurat method
seurat_obj <- ModuleExprScore(
seurat_obj,
n_genes = 25,
method='Seurat'
)
# compute gene scoring for the top 25 hub genes by kME for each module
# with UCell method
library(UCell)
seurat_obj <- ModuleExprScore(
seurat_obj,
n_genes = 25,
method='UCell'
)
```
# Basic Visualization
Here we showcase some of the basic visualization capabilities of scWGCNA,
and we demonstrate how to use some of Seurat's built-in plotting tools to
visualize our scWGCNA results. Note that we have a separate tutorial for visualization of the scWGCNA networks.
## Module Feature Plots
[FeaturePlot](https://www.rdocumentation.org/packages/Seurat/versions/4.1.0/topics/FeaturePlot) is a commonly used Seurat visualization to show a feature of interest
directly on the dimensionality reduction. scWGCNA includes the `ModuleFeaturePlot`
function to consruct FeaturePlots for each co-expression module colored by each
module's uniquely assigned color.
```{r eval=FALSE}
# make a featureplot of hMEs for each module
plot_list <- ModuleFeaturePlot(
seurat_obj,
features='hMEs', # plot the hMEs
order=TRUE # order so the points with highest hMEs are on top
)
# stitch together with patchwork
wrap_plots(plot_list, ncol=6)
```
## Visualization
We can also plot the hub gene signature score using the same function:
scWGCNA includes the `ModuleCorrelogram` function to visualize the correlation between each module based on their hMEs, MEs, or hub gene scores using the R package [corrplot](https://rpubs.com/melike/corrplot).
```{r eval=FALSE}
# plot module correlagram
ModuleCorrelogram(seurat_obj)
```
While scWGCNA includes the `ModuleCorrelogram` function as shown above, we note that R has a rich ecosystem of statistical analysis tools that can be leveraged to understand
the relationship between different modules, and we encourage users to explore
their data beyond the functions that we provide here. Here is an example of
inspecting the correlation structure of the hMEs using an alternative approach,
