Go Back   SEQanswers > Applications Forums > RNA Sequencing

Similar Threads
Thread Thread Starter Forum Replies Last Post
Estimating Dispersions DESeq: ALL Conditions at once, or each comparison separately? carmeyeii Bioinformatics 1 05-20-2013 02:47 AM
Specifing conditions and normaliztion in DESeq? tamari Bioinformatics 0 10-31-2012 02:04 AM
DESeq analysis on subsets of multivariate data jnhutchinson RNA Sequencing 1 06-25-2012 05:39 AM
DESeq: defining conditions mouchkam Bioinformatics 2 04-10-2012 05:45 PM
Is more than two conditions possible in DESEQ? greener RNA Sequencing 5 05-09-2011 04:10 PM

Thread Tools
Old 02-11-2014, 02:36 PM   #1
Junior Member
Location: Texas

Join Date: Feb 2014
Posts: 3
Question DESeq Error -- multivariate conditions with non-multivariate data

I just started doing RNA-seq analysis, and I'm trying to replicate some DESeq results. Long story short, my results were awful, and I can't install the 'pasilla' package (don't ask...) to test DESeq using the vignette.

So...I am now trying to validate my install using the method described in S. Anders et al. Nature Protocols. 2013. In an effort to be brief, I'll only post the results that differed from the paper. If more information is needed, please ask.

I have downloaded the data and compiled everything with 2 notable exceptions:
1) I am using a new build of the D. melanogaster genome (BDGP5.74 instead of BDGP5.70). I don't expect any major issues here.
2) I am using an older version of R, which could be causing problems. Unfortunately, I have not been able to successfully install a newer version due to serious proxy issues.
> sessionInfo()
R version 2.15.0 (2012-03-30)
Platform: x86_64-redhat-linux-gnu (64-bit)
other attached packages:
[1] edgeR_3.0.8         limma_3.14.4        DESeq_1.10.1       
[4] lattice_0.20-6      locfit_1.5-9.1      Biobase_2.18.0     
[7] BiocGenerics_0.4.0  BiocInstaller_1.8.3
I have managed to successfully follow the edgeR analysis without any issues. All of the data matched exactly, with the exception of the following (step 14-B-vi, Use the topTags function to present a tabular summary of the differential expression statistics):
> tt = topTags(de, n = nrow(d)) 
> head(tt$table)
                logFC   logCPM       PValue          FDR
FBgn0039155 -4.646404 5.868219 1.333990e-87 9.599391e-84
FBgn0025111  2.929648 6.855564 1.156645e-54 4.161607e-51
FBgn0003360 -3.227304 8.420833 1.052926e-52 2.525618e-49
FBgn0039827 -3.988696 4.386542 6.595629e-52 1.186554e-48
FBgn0000071  2.718231 4.724487 9.833718e-39 1.415269e-35
FBgn0034736 -3.515391 4.117020 2.467228e-38 2.959028e-35
which had some additional rows than the data given in the paper and different values for the rows that were the same. The next step (vii, Inspect the depth-adjusted reads per million for some of the top differentially expressed genes) had the same extra rows, but the values matched the paper. I imagine these changes are likely due to the newer version of the genome & feature file.

The DESeq analysis goes smoothly (estimating size factors and dispersion) until I run nbinomTest. I am getting the following error:
> res = nbinomTest(cds,"CTL","KD")
Error in nbinomTest(cds, "CTL", "KD") : 
  For CountDataSets with multivariate conditions, only the GLM-based test can be used.
The cds data frame is:
> cds
CountDataSet (storageMode: environment)
assayData: 15682 features, 7 samples 
  element names: counts 
protocolData: none
  sampleNames: CT.SI.1 CT.PA.2 ... CT.SI.7 (7 total)
  varLabels: sizeFactor condition LibraryLayout
  varMetadata: labelDescription
  featureNames: FBgn0000003 FBgn0000008 ... FBgn0264727 (15682 total)
  fvarLabels: disp_pooled
  fvarMetadata: labelDescription
experimentData: use 'experimentData(object)'
> cds$condition
Levels: CTL KD
The samples data frame:
> samples
    LibraryName LibraryLayout
1   Untreated-3        PAIRED
2   Untreated-4        PAIRED
3 CG8144_RNAi-3        PAIRED
4 CG8144_RNAi-4        PAIRED
5   Untreated-1        SINGLE
6 CG8144_RNAi-1        SINGLE
7   Untreated-6        SINGLE
1                                                             SRR031715_1.fastq,SRR031714_1.fastq
2                                                             SRR031717_1.fastq,SRR031716_1.fastq
3                                                             SRR031725_1.fastq,SRR031724_1.fastq
4                                                             SRR031727_1.fastq,SRR031726_1.fastq
5 SRR031713.fastq,SRR031712.fastq,SRR031711.fastq,SRR031710.fastq,SRR031709.fastq,SRR031708.fastq
6 SRR031723.fastq,SRR031722.fastq,SRR031721.fastq,SRR031720.fastq,SRR031719.fastq,SRR031718.fastq
7                                                                 SRR031729.fastq,SRR031728.fastq
                               fastq2 condition shortname              countf
1 SRR031715_2.fastq,SRR031714_2.fastq       CTL   CT.PA.1   Untreated-3.count
2 SRR031717_2.fastq,SRR031716_2.fastq       CTL   CT.PA.2   Untreated-4.count
3 SRR031725_2.fastq,SRR031724_2.fastq        KD   KD.PA.3 CG8144_RNAi-3.count
4 SRR031727_2.fastq,SRR031726_2.fastq        KD   KD.PA.4 CG8144_RNAi-4.count
5                                           CTL   CT.SI.5   Untreated-1.count
6                                            KD   KD.SI.6 CG8144_RNAi-1.count
7                                           CTL   CT.SI.7   Untreated-6.count
which was used to create the deseq data frame:
samplesDESeq = with(samples, 
	data.frame(shortname = I(shortname), 
	countf = I(countf), 
	condition = condition, 
	LibraryLayout = LibraryLayout))
cds = newCountDataSetFromHTSeqCount(samplesDESeq,directory=".")
I have followed the code as provided by the paper, and while I know my way around code, I am new to this pipeline. Am I missing something? Any help would be greatly appreciated! Thanks!
muppetjones is offline   Reply With Quote
Old 02-12-2014, 01:15 AM   #2
Devon Ryan
Location: Freiburg, Germany

Join Date: Jul 2011
Posts: 3,479

You have two types of conditions ("condition" and "LibraryLayout") in samplesDESeq, making it multivariate. Use the GLM methods as the error message suggests.
dpryan is offline   Reply With Quote

deseq, error, r software, rna-seq

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

All times are GMT -8. The time now is 04:23 PM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2018, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO