[BioC] edgeR: estimateGLMCommonDisp - Error in solve.default

[James W. MacDonald]

Hi Tom,

On 10/15/2012 7:52 AM, Tom [guest] wrote:
> Dear list,
>
> when using edgeR GLM after running estimateGLMCommonDisp I am getting this error:
>
> Error in solve.default(R, t(beta)) :
>    system is computationally singular: reciprocal condition number = 1.11022e-16
>
> It seems that there is not enough data for fitting the model (?), therefore I would like to ask you for your help.
>
>
> The design matrix for  (~0 + groups + libraryType) looks like that(2 samples for each group).
>
>     groups0h groups1h groups4h groups24h groups48h libraryTypesingle-end
> 1         1        0        0         0         0                 0
> 2         1        0        0         0         0                 0
> 3         0        1        0         0         0                 0
> 4         0        1        0         0         0                 0
> 5         0        0        1         0         0                 0
> 6         0        0        1         0         0                 0
> 7         0        0        0         1         0                 0
> 8         0        0        0         1         0                 0
> 9         0        0        0         0         1                 1
> 10        0        0        0         0         1                 1

Your design matrix is not full rank, due to the fact that your 48hr 
samples are apparently also the only single-end library. What you are 
trying to do with this design matrix is to estimate the same coefficient 
twice (e.g., the coef for 48h is the same coef for single-end), which 
isn't possible.

Offhand I can only think of three possibilities, all of which are 
suboptimal in some way.

1.) Ignore the fact that the 48h samples are single-end (e.g., remove 
the single-end column from your design matrix).

2.) Omit the 48h sample.

3.) Align all your data as single end, and then fit the model (without 
the single-end coef, obvs).

All three suggestions have drawbacks, and only you can decide which one 
is the most palatable. There may be other possibilities, but none come 
readily to mind. Perhaps others will chime in.

Best,

Jim


>
> Does anyone know how to overcome this problem?
>
> Thank you very much for your help.
> Tom
>
>   -- output of sessionInfo():
>
> R version 2.15.1 (2012-06-22)
> Platform: x86_64-pc-linux-gnu (64-bit)
>
> locale:
>   [1] LC_CTYPE=en_IE.UTF-8       LC_NUMERIC=C               LC_TIME=en_IE.UTF-8        LC_COLLATE=en_IE.UTF-8     LC_MONETARY=en_IE.UTF-8
>   [6] LC_MESSAGES=en_IE.UTF-8    LC_PAPER=C                 LC_NAME=C                  LC_ADDRESS=C               LC_TELEPHONE=C
> [11] LC_MEASUREMENT=en_IE.UTF-8 LC_IDENTIFICATION=C
>
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
>
> other attached packages:
>   [1] BiocInstaller_1.8.2                     biomaRt_2.14.0                          DESeq_1.10.1
>   [4] locfit_1.5-8                            edgeR_3.0.0                             limma_3.14.1
>   [7] TxDb.Hsapiens.UCSC.hg19.knownGene_2.8.0 GenomicFeatures_1.10.0                  AnnotationDbi_1.20.1
> [10] Biobase_2.18.0                          ShortRead_1.16.0                        latticeExtra_0.6-24
> [13] RColorBrewer_1.0-5                      Rsamtools_1.10.1                        lattice_0.20-6
> [16] Biostrings_2.26.2                       GenomicRanges_1.10.2                    IRanges_1.16.2
> [19] BiocGenerics_0.4.0
>
> loaded via a namespace (and not attached):
>   [1] annotate_1.36.0    bitops_1.0-4.1     BSgenome_1.26.1    DBI_0.2-5          genefilter_1.40.0  geneplotter_1.36.0 grid_2.15.1
>   [8] hwriter_1.3        RCurl_1.95-1.1     RSQLite_0.11.2     rtracklayer_1.18.0 splines_2.15.1     stats4_2.15.1      survival_2.36-14
> [15] tools_2.15.1       XML_3.95-0.1       xtable_1.7-0       zlibbioc_1.4.0
>
> --
> Sent via the guest posting facility at bioconductor.org.
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

-- 
James W. MacDonald, M.S.
Biostatistician
University of Washington
Environmental and Occupational Health Sciences
4225 Roosevelt Way NE, # 100
Seattle WA 98105-6099