Unexpected error when using fastMNN

[marionilab]

Unexpected error when using fastMNN

[LTLA]

You're going to have to be more specific. Provide some code to reproduce the error. Also, this question doesn't go here, but on the simpleSingleCell repository (https://github.com/MarioniLab/simpleSingleCell).

[VVictorChen]

Here are the code. I will put the issue on the right place next time. :)

sc4 <- SingleCellExperiment(list(counts=as.matrix(ucb2)),
                           colData=DataFrame(Donor=name4))
#Quality control and normalization
set.seed(1000)
clusters <- quickCluster(sc4, method="igraph", min.mean=0.1)
table(clusters)
sc4 <- computeSumFactors(sc4, min.mean=0.1, clusters=clusters)
summary(sizeFactors(sc4))
sc4 <- normalize(sc4)
vg_seurat <- read.table(file = "*****.txt")
cor_vg <- Reduce(intersect, list(rownames(sc1), rownames(sc2), rownames(sc3), rownames(sc4),vg_seurat$x))
rescaled <- multiBatchNorm(sc1[cor_vg,], sc2[cor_vg,], sc3[cor_vg,], sc4[cor_vg,])
rescaled.sc1 <- rescaled[[1]]
rescaled.sc2 <- rescaled[[2]]
rescaled.sc3 <- rescaled[[3]]
rescaled.sc4 <- rescaled[[4]]
set.seed(100)
original <- list(
    GSEsc1=logcounts(rescaled.sc1)[cor_vg,],
    GSEsc2=logcounts(rescaled.sc2)[cor_vg,],
    GSEsc3=logcounts(rescaled.sc3)[cor_vg,],
    GSEsc4=logcounts(rescaled.sc4)[cor_vg,]
)
mnn.out <- do.call(fastMNN, c(original, list(k=50, d=50, approximate=TRUE, auto.order=TRUE)))

Thanks!

[LTLA]

I'm afraid that your latest post doesn't help; I can hardly reproduce the error when one of the file names is "*****.txt". See https://stackoverflow.com/help/mcve for the expected level of debugging information. The actual error message would also help.

[VVictorChen]

I am sorry for this stupid mistake. The real code is

#input variable genes which are found according to Seurat package
vg_seurat <- read.table(file = "vargene.txt")

[LTLA]

Your are missing the point here. I need a bit of code that I can run on my computer in order to get the same error that you did. Again, see the link to the StackOverflow page that I posted above. You should also describe the error message in more detail than you have currently provided (i.e., none).

[gxg2002cn]

Hi, I met exactly the same error today when running fastMNN. Do you find a solution? Thanks!

[LTLA]

I don't have a solution because I can't reproduce the problem on my computer. I can't even try to reproduce it, because I don't have the data. If I had to guess, there are NAs somewhere in your input data. If you want to help to fix this problem, please provide an RDS file of the input data (anonymized as necessary) and the arguments with which you used to call fastMNN.

[gxg2002cn]

Dear Aaron, Thanks a lot for your reply. I wonder how to run fastMNN without ERCC spick-in even I added? I think my computer setup show be fine because I can go though fastMNN by using example data. But for my data, it stop at multiBatchNorm step. I thought because we used MARS-Seq with ERCC spick-in, but ERCC at quite low level, therefore I limited the gene number to exclude ERCC genes. It worked at this step, but new error came from next step, fastMNN. "BLAS/LAPACK routine DLASCL gave error code -4" as figure attached. I am a biologist and just very beginner of computer programming, I simply put my data in dropbox ( https://www.dropbox.com/sh/chtd75kb4wg5ent/AADtTBmzawGPwVbvc4Wmwe3la?dl=0). Hope you can figure out the problem and please let me know if you need any further information. Thank you very much for your great help. Best regards, Xiaogang [image: image.png] [image: image.png]

…

On Tue, 9 Oct 2018 at 12:57, Aaron Lun ***@***.***> wrote: I don't have a solution because I can't reproduce the problem on my computer. I can't even try to reproduce it, because I don't have the data. If I had to guess, there are NAs somewhere in your input data. If you want to help to fix this problem, please provide an RDS file of the input data (anonymized as necessary) and the arguments with which you used to call fastMNN. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#18 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Aju_EtC-bnwLhwygNnNqN_Hmf6FeRiyBks5ujIErgaJpZM4WgTHb> .

[LTLA]

Thanks @gxg2002cn. As I suspected, you have NaNs in your matrix, corresponding to the spike-ins genes (presumably because you have no counts for the spike-ins). After removing these rows, the function works:

library(scran)
load("fastMNN_error.RData")

set.seed(1000)
not.erccs <- !grepl("ERCC", rownames(original[[1]]))
mnn.out <- do.call(fastMNN, c(original, list(k=20, d=50, approximate=TRUE, 
   subset.row=not.erccs)))

The output looks good, there seems to be decent mixing between batches on a t-SNE plot.

[gxg2002cn]

Great job! It´s working now. Thank you very much Aaron!

…

On Thu, 11 Oct 2018 at 12:35, Aaron Lun ***@***.***> wrote: Thanks @gxg2002cn <https://github.com/gxg2002cn>. As I suspected, you have NaNs in your matrix, corresponding to the spike-ins genes (presumably because you have no counts for the spike-ins). After removing these rows, the function works: library(scran) load("fastMNN_error.RData") set.seed(1000)not.erccs <- !grepl("ERCC", rownames(original[[1]]))mnn.out <- do.call(fastMNN, c(original, list(k=20, d=50, approximate=TRUE, subset.row=not.erccs))) The output looks good, there seems to be decent mixing between batches on a t-SNE plot. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#18 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Aju_EuwmLHprQt2X7uTYCj3yfR91XfkHks5ujx78gaJpZM4WgTHb> .