diff --git a/Tutorial_HTM_2016.Rmd b/Tutorial_HTM_2016.Rmd index d3834bf87f6ba8a48cb1437959896de90ca4a902..e7648741bae01c0b2c8a8c4a7a523e3d091c2b7d 100755 --- a/Tutorial_HTM_2016.Rmd +++ b/Tutorial_HTM_2016.Rmd @@ -165,7 +165,6 @@ sample_n(input_data, 6) ``` - # Using ggplot to create a PCA plot for the data The data we have lives in a ten dimensional space, as every well contains @@ -184,7 +183,6 @@ consists of transforming the counts into percentages by dividing the data for each well by its total number of cells. - ## Grouping, summarizing and data transformation In the code chunk below, we use the`group_by()` function @@ -283,7 +281,7 @@ this plot into an interactive version using `ggplotly` from the `r CRANpkg("plot The first PC nicely separates wells containing various controls from the ones treated with siRNAs. As every component is simply a weighted sum of the original variables, we can inspect these weights (called "loadings") to see which classes -"drive" the components. +"drive" the components and try to interpret what we find. ```{r var_imp, dependson="PCA"} loadings <- PCA$rotation[, 1:2] @@ -309,10 +307,12 @@ define cells that are in mitotic delay / arrest, while the interphase class defi a control category. So a possible explanation for PC1 would be that it separates -cells in mitotic arrest/delay from cells in the interphase -as well as apoptotic cells. (c.f. Figure 1 of @Neumann_2010). - +wells with cells in mitotic arrest/delay (or apoptotic cells) from control +wells with many cells in the interphase phase. (c.f. Figure 1 of @Neumann_2010). +The second principal component seems to separate wells that contain mainly +cells in ana--/metaphase from wells that predominantly contains cells with +strange shape phenotypes. # Plate heatmap of apoptosis proportions diff --git a/Tutorial_HTM_2016.html b/Tutorial_HTM_2016.html index 7332a1b50f11e7dfafc06aaa2b6981dbb4647fb8..1830209d5e57f6c3d1474555c602c3361f2d2be0 100644 --- a/Tutorial_HTM_2016.html +++ b/Tutorial_HTM_2016.html @@ -282,7 +282,7 @@ pl</code></pre></div> </div> <div id="variable-importance-for-the-principal-components" class="section level2"> <h2><span class="header-section-number">7.3</span> Variable importance for the principal components</h2> -<p>The first PC nicely separates wells containing various controls from the ones treated with siRNAs. As every component is simply a weighted sum of the original variables, we can inspect these weights (called “loadingsâ€) to see which classes “drive†the components.</p> +<p>The first PC nicely separates wells containing various controls from the ones treated with siRNAs. As every component is simply a weighted sum of the original variables, we can inspect these weights (called “loadingsâ€) to see which classes “drive†the components and try to interpret what we find.</p> <div class="sourceCode"><pre class="sourceCode r"><code class="sourceCode r">loadings <-<span class="st"> </span>PCA$rotation[, <span class="dv">1</span>:<span class="dv">2</span>] loadings_gg <-<span class="st"> </span>loadings %>% <span class="st"> </span><span class="kw">as.data.frame</span>() %>% @@ -300,7 +300,8 @@ loadings_gg <-<span class="st"> </span>loadings %>% <div class="sourceCode"><pre class="sourceCode r"><code class="sourceCode r"> <span class="co"># geom_hline(aes(yintercept = 0.25, color = I("grey80"))) +</span> <span class="co"># geom_hline(aes(yintercept = -0.25, color = I("coral3")))</span></code></pre></div> <p>We can see that e.g. the “inter†and the “map /prometa†classes as well as the “apo†class are important for PC1. The map and prometa classes combined define cells that are in mitotic delay / arrest, while the interphase class defines a control category.</p> -<p>So a possible explanation for PC1 would be that it separates cells in mitotic arrest/delay from cells in the interphase as well as apoptotic cells. (c.f. Figure 1 of <span class="citation">Neumann et al. (2010)</span>).</p> +<p>So a possible explanation for PC1 would be that it separates wells with cells in mitotic arrest/delay (or apoptotic cells) from control wells with many cells in the interphase phase. (c.f. Figure 1 of <span class="citation">Neumann et al. (2010)</span>).</p> +<p>The second principal component seems to separate wells that contain mainly cells in ana–/metaphase from wells that predominantly contains cells with strange shape phenotypes.</p> </div> </div> <div id="plate-heatmap-of-apoptosis-proportions" class="section level1">