zinbwave Matrix error

[cstrlln]

Hello,
I'm trying to run zinbwave in a cluster and seems to run but I get a Matrix error:

sce_cl2_zinb <- zinbwave(sce_cl2, K = 2, X = "~mouseID + Sex" , epsilon=1e12, BPPARAM=BiocParallel::MulticoreParam(8), verbose = TRUE)
Create model:
ok
Initialize parameters:

ok
Optimize parameters:
Iteration 1
penalized log-likelihood = -21740231789.945
After dispersion optimization = -21739843271.2819

user system elapsed
1.571 9.862 477.966
After right optimization = -10876510949.4719
After orthogonalization = -13745479.7973216
user system elapsed
1.401 13.033 100.598
After left optimization = -12633848.2457174
After orthogonalization = -12410537.8832695
Iteration 2
penalized log-likelihood = -12410537.8832695
After dispersion optimization = -12410537.8832695
user system elapsed
1.155 9.499 134.225
After right optimization = -12379701.8727642
After orthogonalization = -12379673.1041869
user system elapsed
0.900 5.435 37.266
After left optimization = -12378974.9080583
After orthogonalization = -12378974.4038585
Iteration 3
penalized log-likelihood = -12378974.4038585
After dispersion optimization = -12378974.4038585
user system elapsed
0.886 5.043 35.479
After right optimization = -12377788.410242
After orthogonalization = -12377788.3266477
user system elapsed
1.208 7.518 42.181
After left optimization = -12377733.3209309
After orthogonalization = -12377733.3209296
Iteration 4
penalized log-likelihood = -12377733.3209296
After dispersion optimization = -12377733.3209296
user system elapsed
1.067 6.529 49.023
After right optimization = -12377381.0485549
After orthogonalization = -12377381.0388892
user system elapsed
0.975 4.738 55.937
After left optimization = -12377367.3906731
After orthogonalization = -12377367.3906544
Iteration 5
penalized log-likelihood = -12377367.3906544
ok
Error in t.default(x) : argument is not a matrix

Otherwise I'm mostly interested in the weight values for using in edgeR.
please advice.

[jhong26]

I'm also getting a similar error..

[drisso]

Sorry for getting late to this. What version of zinbwave are you using? Can you produce a minimal running example illustrating this behavior?

[jhong26]

Thats for getting back Davide. I am using zinbwave version 1.6.0. My purpose is to use zinbwave for DEseq2 on single cell data from a Seurat object. I have two conditions; I have not added this metadata yet. I just wanted to do a test run first. Here is the output: # sub: Seurat object of single cell data

z <- SingleCellExperiment(assays = list(counts = GetAssayData(object =

sub, slot = "counts")))

z

class: SingleCellExperiment dim: 6537 848 metadata(0): assays(1): counts rownames(6537): 2610015P09Rik AABR07001899.1 ... Zbtb32 Adam18 rowData names(0): colnames(848): SuHx.PH1.SCR3_TGGGATATAAGA SuHx.PH1.SCR3_CCTCGATTAATA ... Control.PH2D2.SCR18_TATACGGGTTGA Control.PH2D2.SCR18_TGTCGTGAAGCC colData names(0): reducedDimNames(0): spikeNames(0):

BiocParallel::register(BiocParallel::SerialParam()) sce_zinb <- zinbwave(z, K = 0, epsilon=1000, observationalWeights = TRUE,

BPPARAM=MulticoreParam(2), verbose=TRUE) #also tried with epsilon=1e13 and got same error Create model: ok Initialize parameters: ok Optimize parameters: Iteration 1 penalized log-likelihood = -1478227.11696393 After dispersion optimization = -1418525.91743817 user system elapsed 0.657 0.386 37.325 After right optimization = -1314985.34430378 After orthogonalization = -1314985.34430378 user system elapsed 0.767 0.548 27.299 After left optimization = -1297233.90997914 After orthogonalization = -1297233.90997914 Iteration 2 penalized log-likelihood = -1297233.90997914 After dispersion optimization = -1297233.90997914 user system elapsed 0.645 0.358 32.728 After right optimization = -1295467.12581696 After orthogonalization = -1295467.12581696 user system elapsed 0.617 0.348 21.994 After left optimization = -1295107.94940988 After orthogonalization = -1295107.94940988 Iteration 3 penalized log-likelihood = -1295107.94940988 After dispersion optimization = -1295107.94940988 user system elapsed 0.642 0.377 30.578 After right optimization = -1295010.39472818 After orthogonalization = -1295010.39472818 user system elapsed 0.590 0.312 22.069 After left optimization = -1294978.88732346 After orthogonalization = -1294978.88732346 Iteration 4 penalized log-likelihood = -1294978.88732346 ok * Error in t.default(x) : argument is not a matrix *

sessionInfo()

R version 3.6.1 (2019-07-05) ... [1] Seurat_3.0.2 biomaRt_2.40.3 ggplot2_3.2.0 [4] magrittr_1.5 scRNAseq_1.10.0 zinbwave_1.6.0 [7] SingleCellExperiment_1.6.0 SummarizedExperiment_1.14.0 DelayedArray_0.10.0 ----- Any input would be much appreciated. Thank you! Jason ...

…

On Mon, Jul 22, 2019 at 11:39 PM Davide Risso ***@***.***> wrote: Sorry for getting late to this. What version of zinbwave are you using? Can you produce a minimal running example illustrating this behavior? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#45?email_source=notifications&email_token=AHVWKG6EZYO6I3H2YICUGG3QA2RQRA5CNFSM4ICRF4BKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2SCTFA#issuecomment-514075028>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHVWKG4QDLOUF2MZVD7I4G3QA2RQRANCNFSM4ICRF4BA> .

[drisso]

Not sure what is going on here. Is counts(z) indeed of class matrix? (Rather than a sparseMatrix or other matrix representations?)

[CUsdevoe]

Was a solution for this error ever discovered? I am encountering the same problem with version 1.10.0

[drisso]

The likely error was that the assay of the SingleCellExperiment object was not of class matrix.

Please, check the class of your input object (e.g., class(x) or class(assay(x))).