Question: DEXSeq error with estimateExonFoldChanges
gravatar for Chen
9 weeks ago by
Chen0 wrote:

Hi there,

I'm try to use DEXSeq to analysis differential exon usage and it worked fine with small sample sizes. But when my sample size goes over 30, the "estimateExonFoldChanges" function start to output the following error:

"Error in .local(x, ...):
 cannot create 126 workers; 125 connections available in this session"

I tried both parallel version (with different number of workers specified) and serial version and met the same problem.

Does anyone know what's causing this error and how to solve it?

BTW, here's my R session info:
R version 3.3.2 (2016-10-31)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: CentOS release 6.7 (Final)

 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8     LC_MONETARY=en_US.UTF-8   
 [6] LC_MESSAGES=en_US.UTF-8    LC_PAPER=en_US.UTF-8       LC_NAME=C                  LC_ADDRESS=C               LC_TELEPHONE=C            

attached base packages:
[1] stats4    parallel  stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] DEXSeq_1.20.2              RColorBrewer_1.1-2         AnnotationDbi_1.36.0       DESeq2_1.14.1              SummarizedExperiment_1.4.0
 [6] GenomicRanges_1.26.1       GenomeInfoDb_1.10.1        IRanges_2.8.1              S4Vectors_0.12.2           Biobase_2.34.0            
[11] BiocGenerics_0.20.0        BiocParallel_1.8.1         ggplot2_2.2.1              dplyr_0.5.0               

loaded via a namespace (and not attached):
 [1] genefilter_1.56.0   statmod_1.4.30      locfit_1.5-9.1      splines_3.3.2       lattice_0.20-34     colorspace_1.3-2    htmltools_0.3.6    
 [8] survival_2.39-5     XML_3.98-1.5        rlang_0.1.1         foreign_0.8-67      DBI_0.5-1           plyr_1.8.4          stringr_1.2.0      
[15] zlibbioc_1.20.0     Biostrings_2.42.1   munsell_0.4.3       gtable_0.2.0        hwriter_1.3.2       memoise_1.0.0       latticeExtra_0.6-28
[22] knitr_1.17          biomaRt_2.30.0      geneplotter_1.52.0  htmlTable_1.7       Rcpp_0.12.11        acepack_1.4.1       xtable_1.8-2       
[29] openssl_0.9.5       scales_0.4.1        base64_2.0          Hmisc_4.0-1         annotate_1.52.0     XVector_0.14.0      Rsamtools_1.26.1   
[36] gridExtra_2.2.1     digest_0.6.12       stringi_1.1.5       grid_3.3.2          tools_3.3.2         bitops_1.0-6        magrittr_1.5       
[43] lazyeval_0.2.0      RCurl_1.95-4.8      tibble_1.3.3        RSQLite_1.1-1       Formula_1.2-1       cluster_2.0.5       Matrix_1.2-7.1     
[50] data.table_1.10.0   assertthat_0.1      R6_2.2.0            rpart_4.1-10        nnet_7.3-12 

Thanks in advance!



I updated R and DEXSeq to the latest version (R 3.4.1, DEXSeq 1.22.0) and still got the same error. Please note that with sample size smaller than 30, the code works perfectly fine.

Here's my code:

dxd <- estimateSizeFactors(dxd)
BPPARAM <- MulticoreParam(workers=detectCores()/2)
formulaFullModel = ~ sample + exon + libType:exon + condition:exon
formulaReducedModel = ~ sample + exon + libType:exon
dxd <- estimateDispersions(dxd, formula=formulaFullModel, BPPARAM=BPPARAM)
dxd_deu <- testForDEU(dxd, reducedModel=formulaReducedModel, fullModel=formulaFullModel, BPPARAM=BPPARAM)
dxd_fc <- estimateExonFoldChanges(dxd_sig)
ADD COMMENTlink modified 8 weeks ago • written 9 weeks ago by Chen0

Hi Chen, 
Have you tried to specify a smaller number of workers?


ADD REPLYlink written 9 weeks ago by Alejandro Reyes1.5k

I tried using serial instead of parallel, and it still produces the same error...

ADD REPLYlink written 8 weeks ago by Chen0

Can you update both R and Bioconductor at least to the latest release versions? If the error still occurs, could you please include the code that reproduces the error messages?

ADD REPLYlink written 8 weeks ago by Alejandro Reyes1.5k

Please see the update in the main question description, thx.

ADD REPLYlink written 8 weeks ago by Chen0

Thanks for adding this. Your note on the sample size makes me think that the problem could be related to RAM usage. What kind of computer are you using?

ADD REPLYlink written 7 weeks ago by Alejandro Reyes1.5k

I feel it's hitting the memory limit. When I specify 350G memory, one experiment with 32 samples successfully finished. But other experiments with more samples failed as usual.

Also, the number of exons matters as well. In one trial, I defined over 100 exons (not real exons) and the program crashed immediately.

ADD REPLYlink written 7 weeks ago by Chen0
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 253 users visited in the last hour