Issue with structure_plot when subsampling and all samples from a group are removed

[eweine]

There appears to be an issue in structure_plot when the n argument is less than the total number of rows of L, and some of the groups that are supplied in the grouping argument have a very small number of samples.

The issue typically occurs here: https://github.com/stephenslab/fastTopics/blob/master/R/structure_plot.R#L256

On line 256, if the length of i is 0, the variable y will not have been created, and this will result in the error

Error in structure_plot(...) : 
  object 'y' not found

I think there are two ways to fix this. Either a) do some sort of "smart" sampling, where you make sure to at least sample 1 row from each group or b) keep the sampling the same but somehow work around this issue.

[pcarbo]

@eweine I ran a small test in which some of the groups disappeared after subsampling, and I was unable to generate an error.
Do you have a small example I could use to reproduce the error? I wonder if perhaps the error only occurs when some combination of the settings are used.

[felixm3]

I have the same issue.

Running this:

structure_plot(fit, 
                             colors = pal, 
                             gap = 25,
                             grouping = seu@meta.data$snn_res.0.3)

Gives this error:

Error in structure_plot(fit, colors = pal, gap = 25, grouping = seu@meta.data$snn_res.0.3): object 'y' not found
Traceback:

1. order(y)
2. .handleSimpleError(function (cnd) 
 . {
 .     watcher$capture_plot_and_output()
 .     cnd <- sanitize_call(cnd)
 .     watcher$push(cnd)
 .     switch(on_error, continue = invokeRestart("eval_continue"), 
 .         stop = invokeRestart("eval_stop"), error = NULL)
 . }, "object 'y' not found", base::quote(structure_plot(fit, colors = pal, 
 .     gap = 25, grouping = seu@meta.data$snn_res.0.3)))

Probably because:

seu@meta.data$snn_res.0.3 %>% table

.
  0   1   2   3   4   5   6 
  0 317 265   0 252 157   9 

Dropping the levels with 0 cells fixes the issue for me:

structure_plot(fit, 
                             colors = pal, 
                             gap = 25,
                             grouping = droplevels(seu@meta.data$snn_res.0.3))

[pcarbo]

Thank you @felixm3! Any chance you can provide the inputs to your structure_plot() call in an .rds or .RData file so I can reproduce the error?

Also, to check, are you using the latest version of fastTopics (version 0.7-38)?