Hierarchical clustering in relative abundances plots

[BrigittevdG]

Hello David,

In the correlation heatmaps from the microViz package, hierarchical clustering using euclidean distances is shown on top/next the heatmap. Is it also possible to include the hierarchical clustering in the relative abundances plots? And would it be possible to get the clusters with the use of the Bray-Curtis dissimilarity (which is already used for the order of the relative abundances plot)?

Kind regards, Brigitte

[david-barnett]

Hi Brigitte,

Not currently possible unfortunately, but could be a nice feature for the future. If you really need it, the code from inside ps_seriate can get you the seriation object which has the clustering dendrogram inside it. You could then plot this alongside the bar chart with ggdendro or dendextend http://andrie.github.io/ggdendro/

(but indeed this would be a bit of manual work)

Code below adapted from inside ps_seriate

method = "OLO_ward"
dist = "bray"
tax_transform = "identity"

# transform taxa for ordering 
ps_transformed <- ps_get(tax_transform(ps, trans = tax_transform))

# calculate distance between samples
distMat <- dist_get(dist_calc(data = ps_transformed, dist = dist))
ser <- seriation::seriate(x = distMat, method = method)

Out of interest, why do you want the clustering dendrogram?

[BrigittevdG]

Hi David, Thanks for the suggestion! I want the clustering to see if there are some groups visible based on the microbiome composition. I am going to try to use your suggestion. Thanks!