<p>Here we will plot the non-linear dimensionality reduction (UMAP) colored by cell type just to check that we have loaded the data correctly, and to make sure that we have grouped cells into clusters and cell types.</p>
<p>In this tutorial, we begin to analyze the co-expresion modules detected in one dataset in external datasets. Rather than building a new co-expression network from scratch in a new dataset, we can take the modules from one dataset and project them into the new dataset. Before starting this tutorial, make sure that you have constructed the co-expression network as in the <ahref="articles/basics_tutorial.html">scWGCNA basics</a>. The main scWGCNA tutorials have been using the control brain samples from <ahref="https://www.nature.com/articles/s41591-019-0695-9"class="external-link">this publication</a>, and now we will project the inhibitory neruon modules from Zhou et al into the control brain samples from <ahref="https://doi.org/10.1038/s41588-021-00894-z"class="external-link">Morabito and Miyoshi 2021</a>.</p>
<p>First we must load the data and the required libraries:</p>
<p>In this section we project the modules from the Zhou et al inhibitory neuron scWGCNA experiment into the Morabito & Miyoshi et al control brain dataset. We refer to the Zhou et al dataset as the “reference” dataset, and the Morabito & Miyoshi et al dataset as the “query” dataset. Just <ahref="basic_tutorial.html">as we had done before</a> when building the co-expression network from scratch, the basic single-cell pipeline has to be 0done on the query dataset (normalization, scaling, variable features, PCA, batch correction, UMAP, clustering). First we make a UMAP plot to visualize the two datasets to ensure they have both been processed.</p>
<spanclass="fu"><ahref="https://ggplot2.tidyverse.org/reference/labs.html"class="external-link">ggtitle</a></span><spanclass="op">(</span><spanclass="st">'Morabito & Miyoshi'</span><spanclass="op">)</span><spanclass="op">+</span>
<p>Next we will run the function <code>ProjectModules</code> to project the modules from the reference dataset into the query dataset. If the genes used for co-expression network analysis in the reference dataset have not been scaled in the query dataset, Seurat’s <code>ScaleData</code> function will be run from within <code>ProjectModules</code>. We perform the following analysis steps to project the modules into the query dataset:</p>
<ul>
<li>Run <code>ProjectModules</code> to compute the module eigengenes in the query dataset based on the gene modules in the reference dataset.</li>
<li>Optionally run <code>ModuleExprScore</code> to compute hub gene expression scores for the projected modules.</li>
<li>Run <code>ModuleConnectivity</code> to compute intramodular connectivity (<em>kME</em>) in the query dataset.</li>
wgcna_name_proj<spanclass="op">=</span><spanclass="st">"Zhou_projected"</span>, <spanclass="co"># name of the new scWGCNA experiment in the query dataset</span>
wgcna_name <spanclass="op">=</span><spanclass="st">"tutorial"</span><spanclass="co"># name of the scWGCNA experiment in the ref dataset</span>
<spanclass="op">)</span></code></pre></div>
<p>As you can see it only takes running one function to project co-expression modules from one dataset into another using scWGCNA. Optionally, we can also compute the hub gene expression scores and the intramodular connectivity for the projected modules.</p>
