Plot proportion of cells per level as label

[hummuscience]

I was wondering how I could plot a label that contains the proportion of cells in the current cluster by resolution/clustering depth.

[lazappi]

Hi @Cumol

Thanks for giving {clustree} a go! If I am understanding you correctly you want to add a label with the proportion of the total samples that belong to each cluster (something like cluster_size / total_samples)? Every sample should be present at each resolution so total_samples should be the same whichever resolution you look at.

This sounds like something that should be easy to do but unfortunately I don't think that it is possible at the moment. It is easy to add the cluster size as a label:

clustree(iris_clusts, prefix = "K", node_label = "size")

But I don't think it is possible to do a proportion due to the way the node aggregation functions work. I think I might have a way to make this possible but I need to try some things and see if they work. I will comment here if I am able to come up with a solution.

Thanks for raising the issue 🎉! I hadn't considered this a problem before and I think it's functionality we should have 😸.

[hummuscience]

Thanks for the quick reply!

I found the solution for "size" and if I understood correctly, "size" is calculated internally, which is why you can pass it as a node label. I was wondering whether the total_samples is also calculated internally and it would be possible to do node_label = "size/total_samples".

Or allow general access to "resolution statistics". I saw another issue where someone calculated the statistics beforehand and then added the values where needed, but that looked too complex for me :D

[lazappi]

Yeah, unfortunately that won't work the way things are set up at the moment. I think the workaround I have thought will make that possible but I will have to try and implement it to see.

[lazappi]

Ok, my solution didn't work quite as well as I had hoped but I have come up with a slightly hacky workaround for your specific case. Try something like this and see if it gives you what you want.

# You need to replace 150 here with however many samples you have
# I think the only way to do this is to hard code it for now
prop_samples <- function(x) {length(x) / 150}

clustree(YOUR_DATA, prefix = "YOUR_PREFIX", node_label = "ANY_METADATA_COLUMN", 
         node_label_aggr = "prop_samples")

[hummuscience]

Will give it a try and let you know.

[lazappi]

@Cumol Was this helpful in the end?

[plijnzaad]

I'm not Cumol but yes this works very nicely, I'm using , node_label_aggr='n_cells', with
n_cells <- function(x)sprintf("%g cells (%.1f%%)", length(x), 100*length(x)/nrow(df))
where df is the data.frame that is fed to clustree. Many thanks!

PS: I was hoping to be able to say edge_color="black" (it confuses my users) but that doesn't work for some reason?

[lazappi]

PS: I was hoping to be able to say edge_color="black" (it confuses my users) but that doesn't work for some reason?

Please open a new issue if this is a feature you would like to have. At the moment there is no parameter for controlling the edge colour.