I was running a pipeline which I developed some time ago and we spotted a very weird behaviour that had quite an impact over our analyses.
Indeed, the scater's runTSNE() function was subjected to a change in the parameters names and, in particular, the old use_dimred parameter now has a new name, which is dimred.
Interestingly, I was running the pipeline using the runTSNE() function and passing the dimensionality reduction through use_dimred which is apparently deprecated and obtaining very weird results without being aware of that since no warning and/or error was returned.
A quick example:
# load some data
library("scRNAseq")
library("scater")
sce.416b <- LunSpikeInData(which="416b")
check the data and notice no dimensionality reduction is available:
> sce.416b
class: SingleCellExperiment
dim: 46604 192
metadata(0):
assays(1): counts
rownames(46604): ENSMUSG00000102693 ENSMUSG00000064842 ...
ENSMUSG00000095742 CBFB-MYH11-mcherry
rowData names(1): Length
colnames(192): SLX-9555.N701_S502.C89V9ANXX.s_1.r_1
SLX-9555.N701_S503.C89V9ANXX.s_1.r_1 ...
SLX-11312.N712_S508.H5H5YBBXX.s_8.r_1
SLX-11312.N712_S517.H5H5YBBXX.s_8.r_1
colData names(9): Source Name cell line ... spike-in addition block
reducedDimNames(0): <-- HERE IT IS EMPTY
mainExpName: endogenous
altExpNames(2): ERCC SIRV
at this point I simply run the runTSNE() function using a non-existent dimred:
# add logcounts
sce.416b <- logNormCounts(sce.416b)
# add random colours to cells
colLabels(sce.416b) <- rep(c(1,2,3,4), 48)
# runTSNE, first try
sce.416b <- runTSNE(sce.416b, use_dimred="made_up_dimred", perplexity=30)
# and plot
dev.new(); plotTSNE(sce.416b, colour_by="label")
# runTSNE, second try
sce.416b <- runTSNE(sce.416b, use_dimred="another_made_up_dimred", perplexity=30)
# and plot
dev.new(); plotTSNE(sce.416b, colour_by="label")
# runTSNE, third try
sce.416b <- runTSNE(sce.416b, use_dimred="a_third_made_up_dimred", perplexity=30)
# and plot
dev.new(); plotTSNE(sce.416b, colour_by="label")
it, unfortunately, works. Indeed, I now have three TSNE available that I did plot and that are slightly different with each other.
> sce.416b
class: SingleCellExperiment
dim: 46604 192
metadata(0):
assays(2): counts logcounts
rownames(46604): ENSMUSG00000102693 ENSMUSG00000064842 ...
ENSMUSG00000095742 CBFB-MYH11-mcherry
rowData names(1): Length
colnames(192): SLX-9555.N701_S502.C89V9ANXX.s_1.r_1
SLX-9555.N701_S503.C89V9ANXX.s_1.r_1 ...
SLX-11312.N712_S508.H5H5YBBXX.s_8.r_1
SLX-11312.N712_S517.H5H5YBBXX.s_8.r_1
colData names(10): Source Name cell line ... block sizeFactor
reducedDimNames(1): TSNE <-- THERE YOU GO: A NICE TSNE COMPUTED OVER A NON-EXISTENT DIMRED
mainExpName: endogenous
altExpNames(2): ERCC SIRV
What is even more interesting is that if I run the function with the newly renamed parameter, i.e. dimred, I do get an error:
> sce.416b <- runTSNE(sce.416b, dimred="made_up_dimred", perplexity=30)
Error in value[[3L]](cond) :
invalid subscript 'type' in 'reducedDim(<SingleCellExperiment>, type="character", ...)':
'made_up_dimred' not in 'reducedDimNames(<SingleCellExperiment>)'
But then, if you have an old pipeline which used the runTSNE function with the use_dimred and you're now running an updated version of scater, I have no idea what you may get.
So, as per my question...is this a bug? Has anyone else noticed this weird behaviour? Am I wrong?
