Question

DESeq2: Error in cleanContrast

0

Entering edit mode

pokharel.kisun • 0

@pokharelkisun-9612

Last seen 5.7 years ago

Hi,
I have sample table that looks like below:

>head(sampletable)
                  SampleId SampleName Location Tissue
1 FR10_MCRP_374D_S108_L005     FR10MC       FR    MC
2  FR10_MCRP_374D_S56_L001     FR10MC       FR    MC
3  FR10_MCRP_374D_S56_L002     FR10MC       FR    MC
4  FR10_MCRP_374D_S56_L003     FR10MC       FR    MC
5  FR10_MCRP_374D_S56_L004     FR10MC       FR    MC
6  FR10_MCRP_374D_S56_L005     FR10MC       FR    MC

The data comes from two locations (FR and YR) and four tissues (MC, PR, SC and TH).

>sampletable$Location
  [1] FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR
 [59] FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR FR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR
[117] YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR YR
Levels: FR YR

> sampletable$Tissue
  [1] MC  MC  MC  MC  MC  MC  MC  PR  PR  PR  PR  PR  PR  PR  SC  SC  SC  SC  SC  SC  SC  TH  TH  TH  TH  TH  TH  TH  MC  MC  MC  MC  MC  MC  MC  PR  PR  PR  PR  PR  PR  PR  SC  SC 
 [45] SC  SC  SC  SC  SC  TH  TH  TH  TH  TH  TH  TH  MC  MC  MC  MC  MC  MC  MC  PR  PR  PR  PR  PR  PR  PR  SC  TH  TH  TH  TH  TH  TH  TH  MC  MC  MC  MC  MC  MC  MC  PR  PR  PR 
 [89] PR  PR  PR  PR  SC  TH  TH  TH  TH  TH  TH  TH  MC  MC  MC  MC  MC  MC  MC  PR  PR  PR  PR  PR  PR  PR  SC  SC  SC  SC  SC  SC  SC  TH  TH  TH  TH  TH  TH  TH  MC  MC  MC  MC 
[133] MC  MC  MC  PR  PR  PR  PR  PR  PR  PR  SC  SC  SC  SC  SC  SC  SC  TH  TH  TH  TH  TH  TH  TH 
Levels: MC  PR SC TH

I want to make differential expression between two locations for each tissues so I combined Location and Tissue to form group and ran deseq2.

> dds$group
  [1] FRMC  FRMC  FRMC  FRMC  FRMC  FRMC  FRMC  FRPR  FRPR  FRPR  FRPR  FRPR  FRPR  FRPR  FRSC  FRSC  FRSC  FRSC  FRSC  FRSC  FRSC  FRTH  FRTH  FRTH  FRTH  FRTH  FRTH  FRTH  FRMC 
 [30] FRMC  FRMC  FRMC  FRMC  FRMC  FRMC  FRPR  FRPR  FRPR  FRPR  FRPR  FRPR  FRPR  FRSC  FRSC  FRSC  FRSC  FRSC  FRSC  FRSC  FRTH  FRTH  FRTH  FRTH  FRTH  FRTH  FRTH  FRMC  FRMC 
 [59] FRMC  FRMC  FRMC  FRMC  FRMC  FRPR  FRPR  FRPR  FRPR  FRPR  FRPR  FRPR  FRSC  FRTH  FRTH  FRTH  FRTH  FRTH  FRTH  FRTH  YRMC  YRMC  YRMC  YRMC  YRMC  YRMC  YRMC  YRPR  YRPR 
 [88] YRPR  YRPR  YRPR  YRPR  YRPR  YRSC  YRTH  YRTH  YRTH  YRTH  YRTH  YRTH  YRTH  YRMC  YRMC  YRMC  YRMC  YRMC  YRMC  YRMC  YRPR  YRPR  YRPR  YRPR  YRPR  YRPR  YRPR  YRSC  YRSC 
[117] YRSC  YRSC  YRSC  YRSC  YRSC  YRTH  YRTH  YRTH  YRTH  YRTH  YRTH  YRTH  YRMC  YRMC  YRMC  YRMC  YRMC  YRMC  YRMC  YRPR  YRPR  YRPR  YRPR  YRPR  YRPR  YRPR  YRSC  YRSC  YRSC 
[146] YRSC  YRSC  YRSC  YRSC  YRTH  YRTH  YRTH  YRTH  YRTH  YRTH  YRTH 
Levels: FRMC  FRPR FRSC FRTH YRMC  YRPR YRSC YRTH

>resultsNames(dds)

[1] "Intercept"            "group_FRPR_vs_FRMC."  "group_FRSC_vs_FRMC."  "group_FRTH_vs_FRMC."  "group_YRMC._vs_FRMC." "group_YRPR_vs_FRMC."  "group_YRSC_vs_FRMC." 
[8] "group_YRTH_vs_FRMC."

Do you know why there is "." with FRMC and why this design does not work?

I get the following error:

> MC_YR-FR<-results(dds, contrast = c("group", "YRMC", "FRMC"))
Error in cleanContrast(object, contrast, expanded = isExpanded, listValues = listValues,  : 
  YRMC and FRMC should be levels of group such that group_YRMC_vs_FRMC. and group_FRMC_vs_FRMC. are contained in 'resultsNames(object)'

Thanks.

deseq2 design and contrast matrix • 6.6k views

ADD COMMENT • link updated 6.8 years ago by Peter Langfelder ★ 3.0k • written 6.8 years ago by pokharel.kisun • 0

1

Entering edit mode

Michael Love 41k

@mikelove

Last seen 6 hours ago

United States

I can't see why that would happen. Can you provide all your code and your sessionInfo()?

ADD COMMENT • link 6.8 years ago Michael Love 41k

0

Entering edit mode

Hi Mike, the later error also had to do with the symbol "-". Once I changed it to MC_FR_MR, it didn't throw any error message.

ADD REPLY • link 6.8 years ago pokharel.kisun • 0

score 3 · Accepted Answer · 2017-07-12

3

Entering edit mode

Peter Langfelder ★ 3.0k

@peter-langfelder-4469

Last seen 22 days ago

United States

Your "MC" level of tissue is most likely actually "MC ". Note the space at the end. The telltale sign of this is the apparently uneven spacing of the levels in your outputs. The space is converted to a period (".") when column names are checked deep inside the DESeq2 analysis pipeline. Make sure the MC and other levels do not contain spaces, or you will be forever fighting spaces turning into periods and running into issues like you did above.

HTH,

Peter

ADD COMMENT • link 6.8 years ago Peter Langfelder ★ 3.0k

0

Entering edit mode

Thanks Peter!!

ADD REPLY • link 6.8 years ago Michael Love 41k

0

Entering edit mode

Glad to be of help. Although this is not your (or DESeq2's) fault at all, it may be helpful to add some sort of a check inside your pipeline that factor levels are valid names. This type of a problem (factor levels that are not valid names) is probably quite common but it can be difficult for non-R-programmers to figure out what's going on and how to fix it. Been there, done that...

ADD REPLY • link 6.8 years ago Peter Langfelder ★ 3.0k

0

Entering edit mode

Currently the only check is that, after conversion by make.names(), which is required for other R code to run, that the levels are still unique.

But I could just as well put a check that levels are the same before and after make.names(), so people don't get confused with how things are renamed by R. Can't think of a reason why not to add that check. Thanks for the suggestion.

ADD REPLY • link 6.8 years ago Michael Love 41k

0

Entering edit mode

Fixed. This could potentially throw some errors for existing code out there (it did in vignette, where i had to change single-read and paired-end to single and paired), but I'd prefer to help users avoid confusion like above.

Numbers, while not being valid R names, are perfectly safe as DESeq2 factor levels, so I instead insist that the levels are just alphanumeric plus '_' or '.'

New behavior is:

 > counts <- matrix(1:16,ncol=4)
 > coldata <- data.frame(x=factor(c("A","A ","B","B")))
!> dds <- DESeqDataSetFromMatrix(counts, coldata, ~x)
 Error in DESeqDataSet(se, design = design, ignoreRank) :
   levels of factors in the design contain characters other than letters, numbers, '_' and '.'

ADD REPLY • link 6.8 years ago Michael Love 41k

0

Entering edit mode

I switched this to a message rather than a warning, because it's not necessary to break code in order to give users the message that they have e.g. spaces in their factor levels

https://github.com/Bioconductor-mirror/DESeq2/commit/b76d4f83fe835ad91d9fad87478ca0f0bf3a4212

ADD REPLY • link 6.7 years ago Michael Love 41k