search

kharchenkolab / numbat

Haplotype-aware CNV analysis from single-cell RNA-seq
https://kharchenkolab.github.io/numbat/
Other
163 stars 23 forks source link

Subtrees disappear #150

Open acalcino opened 9 months ago

acalcino commented 9 months ago

Hello, I have been using Numbat for a while but recently did a reinstall. Since then my results seem to be a bit off and I noticed in the stderr I get this issue:

running hclust... ! # Invaild edge matrix for . A is returned. ! # Invaild edge matrix for . A is returned.

I googled it and it looks like an issue with tidytree which according to this:

https://support.bioconductor.org/p/9153297/#9153321

Can be fixed by reverting back to v0.4.2 however when I do this, I upset numbat because it wants v0.4.5. I also noticed this issue in the output presented in issue #141 but it wasn't spoken about.

teng-gao commented 9 months ago

Thanks for reporting this. Maybe it's an object conversion problem between phylo and tidytree with the newer version. How do the results look different?

acalcino commented 9 months ago

Here's an example. The first plot (bulk_clones_final) was made before I did the reinstall and its beautiful (numbat is awesome btw!). The second was created using the exact same parameters but it gave this error I mentioned and you can see that it fails to find any of the previous cnvs.

bulk_clones_final bulk_clones_final

You can see too that the number of cells in the second one is substantially lower than the first even though they both started with the same inputs.

teng-gao commented 9 months ago

Hi @acalcino ,

Thanks. I indeed see the same warning message "! # Invaild edge matrix for . A is returned." when I upgrade to tidytree v0.4.5. However my results are the same when running on test cases. Would you share your input data via email so I can take a look?

Thanks, Teng

teng-gao commented 9 months ago

Also, does this happen on all your samples or just one?

acalcino commented 8 months ago

Hello again, Sorry for the late reply. I was away for a conference for a week, then we got hit by a cyclone, then the city flooded and now the highways to the uni campus are closed due to damage so I've been 'working from home' with two kids and torrential rain outside.

Anyway, I've only tried a before and after on that one sample but I do get that error every time I try and run the new version of Numbat. The samples I processed with the new Numbat all gave strange/unexpected results but as I don't have a control for them, I can't say for sure if this is related to the upgrade or whether its related to the samples themselves.

teng-gao commented 8 months ago

Hi @acalcino,

Wow, hope you and the family are ok.

What you describe indeed sounds abnormal. Let me know if you’re ok with sharing the input data for one of these examples privately (you can use my email tgao@g.harvard.edu) so that I can debug this.

teng-gao commented 6 months ago

The invalid edge matrix warning is fixed in 7e936c30bcadf319b38302b107da8f82cb97e174 The actual difference in results is actually caused by an incompatibility of igraph 2.0.x with tidygraph 1.3.1 .. see https://github.com/thomasp85/tidygraph/issues/190 A fix was introduced in Numbat v1.4.0