[BioC] Problems with edgeR for differential expression

Nick Schurch N.Schurch at dundee.ac.uk
Tue Oct 11 16:19:10 CEST 2011


Hi Gordon,

Thanks for the quick response, and sorry I didn't get back to you yesterday.
I'll try to address each of your questions clearly.

1)

>> Finally, would you be willing to share some of your date with us offline
so that we can >> trouble-shoot?  As I said, we haven't seen this behaviour.

We'd be happy to share the anonymized data with you offline. Giving you the
DGElist, or the dataframe would be best from our perspective. What would be
the best way of getting it to you?

2)

>> Can you try updating to edgeR 2.3.52 (the current devel version) to see
it makes a
>> difference?  (There are many changes, but one is that exactTest() is much
faster now.)

I have downloaded v2.3.52, but I'm having some problems building it.

R CMD INSTALL edgeR_2.3.52.tar.gz
* installing to library ‘/homes/nschurch/myRlibrary/x86_64’
* installing *source* package ‘edgeR’ ...
** R
** data
** inst
** preparing package for lazy loading
Error : class "MDS" is not exported by 'namespace:limma'
ERROR: lazy loading failed for package ‘edgeR’
* removing ‘/homes/nschurch/myRlibrary/x86_64/edgeR’

I have limma (v3.4.4) installed and it seems to be working fine so I'm not
quite sure what is going on here. Do I need a more up to date version of
limma for the development build of edgeR?

3)

>> Given your experimental layout with multiple groups but no covariates,
you could use
>> the "classic" edgeR functions:
>>
>> estimateCommonDisp(y)
>> exactTest()
>> topTags()
>>
>> rather than the glm code.  This would allow you to compare any two of
your groups.
>> What results does this give?

This seems to give sensible results; I don't get any NaN fold-changes. But
it seems to present some new questions. A quick check confirms that the
fold-changes calculated by each method are tightly correlated (for those
that aren't 'NaN'), but a plot of the p-values calculated by each method is
more complicated.

Attached are two .pngs of this plot, one for the full range of p-values, and
one for a zoomed in region of the plot. In these figures, the blue points
are genes with non-NaN fold-changes in both calculations and the red points
are those with NaN fold-changes in the glm calculation. The solid red line
marks the 1:1 line.

While the p-values are correlated, there is considerable structure to the
plot. There seems to be at least 3 distinct regions to the plot that are al
removed from the 1:1 line. The zoomed plot also shows clear discretization
of the p-values from the exactTest for the red points (and maybe also for
the glm p-values), that isn't clearly seen in the blue points. This makes me
think that while the exactTest is calculating a reasonable-looking fold
change, that they are not to be trusted and that something funny is going
on. We are also getting strange results when comparing edgeR p-values (from
either method) with those calculated by DESeq for the same data (I can send
you these plots if you're interested).

I really think there is something funny going on here. Have you seen
anything like these structures when comparing the p-values calculated by
each method?

4)

>> Thanks for including output from your objects, but I find that I can't
get much information >> from the str(y) output.  Could you please use the
show method for these objects, e.g.,
>> show(y)?

> # create the design matrix
> model.groups<-groups
> model.factors<-as.factor(model.groups)
> model<-model.matrix(~model.factors)
>
> # build DGElist and calculate normalization factors
> x=as.data.frame(data)
> rownames(x)=genenames
> y=DGEList(x,group=groups)
> y=calcNormFactors(y)
> show(y)

An object of class "DGEList"
$samples
                group lib.size norm.factors
cond07.rep0020 cond07  1866963    0.9916452
cond07.rep0021 cond07  2364994    0.9783033
cond07.rep0022 cond07  1712838    1.0023115
cond08.rep0023 cond08  2782920    1.0357778
cond08.rep0025 cond08  2780054    0.9724019
12 more rows ...

$counts
      cond07.rep0020 cond07.rep0021 cond07.rep0022 cond08.rep0023
38536              7              6              7             11
38537            187            169            146            265
38538             44            117             78             78
38539              0             23             22             40
38540             98            144             97            215
      cond08.rep0025 cond08.rep0024 cond07.rep0019 cond07.rep0018
38536              7              9              5              9
38537            186            279            237            132
38538            160            109              7             73
38539             14             23             11             13
38540            175            247            153             87
      cond00.rep0029 cond00.rep0028 cond06.rep0015 cond06.rep0016
38536              0             10              5             18
38537            277            153            150            240
38538             46             52             62             84
38539             23             21             13             15
38540            141            119             79            211
      cond06.rep0017 cond08.rep0026 cond10.rep0030 cond10.rep0031
38536              0              8              5             14
38537             93            282            211            258
38538             21            120            100             83
38539              9             30             34             42
38540             46            233            208            223
      cond00.rep0027
38536              7
38537            238
38538             47
38539             21
38540            231
6346 more rows ...

$all.zeros
38536 38537 38538 38539 38540
FALSE FALSE FALSE FALSE FALSE
6346 more elements ...

> # estimate dispersion and fit models
> z=estimateGLMCommonDisp(y, design))
> show(z)

An object of class "DGEList"
$samples
                group lib.size norm.factors
cond07.rep0020 cond07  1866963    0.9916452
cond07.rep0021 cond07  2364994    0.9783033
cond07.rep0022 cond07  1712838    1.0023115
cond08.rep0023 cond08  2782920    1.0357778
cond08.rep0025 cond08  2780054    0.9724019
12 more rows ...

$counts
      cond07.rep0020 cond07.rep0021 cond07.rep0022 cond08.rep0023
38536              7              6              7             11
38537            187            169            146            265
38538             44            117             78             78
38539              0             23             22             40
38540             98            144             97            215
      cond08.rep0025 cond08.rep0024 cond07.rep0019 cond07.rep0018
38536              7              9              5              9
38537            186            279            237            132
38538            160            109              7             73
38539             14             23             11             13
38540            175            247            153             87
      cond00.rep0029 cond00.rep0028 cond06.rep0015 cond06.rep0016
38536              0             10              5             18
38537            277            153            150            240
38538             46             52             62             84
38539             23             21             13             15
38540            141            119             79            211
      cond06.rep0017 cond08.rep0026 cond10.rep0030 cond10.rep0031
38536              0              8              5             14
38537             93            282            211            258
38538             21            120            100             83
38539              9             30             34             42
38540             46            233            208            223
      cond00.rep0027
38536              7
38537            238
38538             47
38539             21
38540            231
6346 more rows ...

$all.zeros
38536 38537 38538 38539 38540
FALSE FALSE FALSE FALSE FALSE
6346 more elements ...

$common.dispersion
[1] 0.1012043

> fit<-glmFit(z,model,dispersion=z$common.dispersion)
> show(fit)

An object of class "DGEGLM"
$coefficients
      (Intercept) model.factorscond06 model.factorscond07
model.factorscond08
38536  -12.885005          0.48913051           0.2967643
0.1212677129
38537   -9.219056         -0.01086056          -0.1339230
-0.1877871229
38538  -10.724813          0.38699166           0.3731979
0.5624151466
38539  -11.535562         -0.22950901          -0.3546394
-0.1163631383
38540   -9.555217         -0.13261443          -0.2158204
-0.0002525462
      model.factorscond10
38536           0.1592107
38537          -0.2828476
38538           0.2987787
38539           0.2134461
38540          -0.0250628
6346 more rows ...

$df.residual
[1] 12 12 12 12 12
6346 more elements ...

$deviance
    38536     38537     38538     38539     38540
23.297974  3.448158 30.568649 25.886776  3.073930
6346 more elements ...

$design
  (Intercept) model.factorscond06 model.factorscond07 model.factorscond08
1           1                   0                   1                   0
2           1                   0                   1                   0
3           1                   0                   1                   0
4           1                   0                   0                   1
5           1                   0                   0                   1
  model.factorscond10
1                   0
2                   0
3                   0
4                   0
5                   0
12 more rows ...

$offset
         [,1]     [,2]     [,3]     [,4]     [,5]     [,6]     [,7]     [,8]
[1,] 14.43143 14.65435 14.35597 14.87416 14.80999 14.96341 14.60526 14.49537
[2,] 14.43143 14.65435 14.35597 14.87416 14.80999 14.96341 14.60526 14.49537
[3,] 14.43143 14.65435 14.35597 14.87416 14.80999 14.96341 14.60526 14.49537
[4,] 14.43143 14.65435 14.35597 14.87416 14.80999 14.96341 14.60526 14.49537
[5,] 14.43143 14.65435 14.35597 14.87416 14.80999 14.96341 14.60526 14.49537
         [,9]    [,10]    [,11]    [,12]    [,13]   [,14]    [,15]    [,16]
[1,] 14.56309 14.43248 14.25811 14.76053 13.69444 15.0837 14.85685 15.05181
[2,] 14.56309 14.43248 14.25811 14.76053 13.69444 15.0837 14.85685 15.05181
[3,] 14.56309 14.43248 14.25811 14.76053 13.69444 15.0837 14.85685 15.05181
[4,] 14.56309 14.43248 14.25811 14.76053 13.69444 15.0837 14.85685 15.05181
[5,] 14.56309 14.43248 14.25811 14.76053 13.69444 15.0837 14.85685 15.05181
        [,17]
[1,] 14.85989
[2,] 14.85989
[3,] 14.85989
[4,] 14.85989
[5,] 14.85989
6346 more rows ...

$samples
                group lib.size norm.factors
cond07.rep0020 cond07  1866963    0.9916452
cond07.rep0021 cond07  2364994    0.9783033
cond07.rep0022 cond07  1712838    1.0023115
cond08.rep0023 cond08  2782920    1.0357778
cond08.rep0025 cond08  2780054    0.9724019
12 more rows ...

$genes
NULL

$dispersion
[1] 0.1012043

$lib.size
 [1] 1851365.0 2313681.4 1716797.3 2882486.8 2703329.9 3151566.4 2202855.3
 [8] 1973607.7 2111876.1 1853304.7 1556748.9 2572866.3  885968.3 3554417.9
[15] 2833000.8 3442857.6 2841627.6

$weights
NULL

$fitted.values
      cond07.rep0020 cond07.rep0021 cond07.rep0022 cond08.rep0023
38536       6.316675       7.894054       5.857544       8.251761
38537     160.525760     200.611698     148.857840     236.824760
38538      59.134748      73.901673      54.836500     111.248471
38539      12.695304      15.865531      11.772537      25.084212
38540     105.677922     132.067448      97.996653     204.117136
      cond08.rep0025 cond08.rep0024 cond07.rep0019 cond07.rep0018
38536       7.738885       9.022061       7.515926       6.733756
38537     222.105256     258.932306     191.002330     171.125028
38538     104.333981     121.633494      70.361758      63.039324
38539      23.525139      27.425819      15.105567      13.533555
38540     191.430528     223.171432     125.741373     112.655672
      cond00.rep0029 cond00.rep0028 cond06.rep0015 cond06.rep0016
38536       5.355276       4.699593       6.438114       10.64038
38537     209.354902     183.722151     152.656946      252.29883
38538      46.445257      40.758647      50.415007       83.32177
38539      20.646073      18.118233      12.097986       19.99456
38540     149.585764     131.270957      96.570986      159.60457
      cond06.rep0017 cond08.rep0026 cond10.rep0030 cond10.rep0031
38536       3.664024       10.17531        8.42373       10.23710
38537      86.879281      292.03054      211.65193      257.21400
38538      28.691911      137.18139       83.99967      102.08218
38539       6.885139       30.93155       34.28579       41.66645
38540      54.959948      251.69850      195.69687      237.82431
      cond00.rep0027
38536       7.205773
38537     281.696762
38538      62.494255
38539      27.780252
38540     201.274606
6346 more rows ...

$abundance
     38536      38537      38538      38539      38540
-12.665286  -9.334396 -10.362213 -11.655346  -9.639765
6346 more elements ...

> # liklihood ratio statistics
> results=glmLRT(z, fit, coef = 4)
> show(results)

An object of class "DGELRT"
$samples
                group lib.size norm.factors
cond07.rep0020 cond07  1866963    0.9916452
cond07.rep0021 cond07  2364994    0.9783033
cond07.rep0022 cond07  1712838    1.0023115
cond08.rep0023 cond08  2782920    1.0357778
cond08.rep0025 cond08  2780054    0.9724019
12 more rows ...

$all.zeros
38536 38537 38538 38539 38540
FALSE FALSE FALSE FALSE FALSE
6346 more elements ...

$common.dispersion
[1] 0.1012043

$table
         logConc       logFC LR.statistic   p.value
38536 -12.665286  0.70566616  1.386712521 0.2389611
38537  -9.334396 -0.01566848  0.001650326 0.9675955
38538 -10.362213  0.55831094  1.833689742 0.1756924
38539 -11.655346 -0.33111151  0.467320591 0.4942224
38540  -9.639765 -0.19132218  0.240419682 0.6239032
6346 more rows ...

$coefficients.full
      (Intercept) model.factorscond06 model.factorscond07
model.factorscond08
38536  -12.885005          0.48913051           0.2967643
0.1212677129
38537   -9.219056         -0.01086056          -0.1339230
-0.1877871229
38538  -10.724813          0.38699166           0.3731979
0.5624151466
38539  -11.535562         -0.22950901          -0.3546394
-0.1163631383
38540   -9.555217         -0.13261443          -0.2158204
-0.0002525462
      model.factorscond10
38536           0.1592107
38537          -0.2828476
38538           0.2987787
38539           0.2134461
38540          -0.0250628
6346 more rows ...

$coefficients.null
      (Intercept) model.factorscond07 model.factorscond08
model.factorscond10
38536  -12.627557          0.03931612         -0.13618044
-0.09823747
38537   -9.224403         -0.12857630         -0.18244047
-0.27750097
38538  -10.518433          0.16681834          0.35603561
0.09239918
38539  -11.634560         -0.25564159         -0.01736528
0.31244392
38540   -9.618361         -0.15267688          0.06289095
0.03808070
6346 more rows ...

$design.full
  (Intercept) model.factorscond06 model.factorscond07 model.factorscond08
1           1                   0                   1                   0
2           1                   0                   1                   0
3           1                   0                   1                   0
4           1                   0                   0                   1
5           1                   0                   0                   1
  model.factorscond10
1                   0
2                   0
3                   0
4                   0
5                   0
12 more rows ...

$design.null
  (Intercept) model.factorscond07 model.factorscond08 model.factorscond10
1           1                   1                   0                   0
2           1                   1                   0                   0
3           1                   1                   0                   0
4           1                   0                   1                   0
5           1                   0                   1                   0
12 more rows ...

$dispersion.used
[1] 0.1012043

$comparison
[1] "model.factorscond06"


I hope this helps!

Nick



On 8 October 2011 02:02, Gordon K Smyth <smyth at wehi.edu.au> wrote:

> Hi Nick,
>
> We haven't seen anything like this.  Here are some suggestions.
>
> Given your experimental layout with multiple groups but no covariates, you
> could use the "classic" edgeR functions:
>
>  estimateCommonDisp(y)
>  exactTest()
>  topTags()
>
> rather than the glm code.  This would allow you to compare any two of your
> groups.  What results does this give?
>
> Can you try updating to edgeR 2.3.52 (the current devel version) to see it
> makes a difference?  (There are many changes, but one is that exactTest() is
> much faster now.)
>
> Thanks for including output from your objects, but I find that I can't get
> much information from the str(y) output.  Could you please use the show
> method for these objects, e.g., show(y)?
>
> Finally, would you be willing to share some of your date with us offline so
> that we can trouble-shoot?  As I said, we haven't seen this behaviour.
>
> Best wishes
> Gordon
>
>  Date: Fri, 7 Oct 2011 10:10:21 +0100
>> From: Nick Schurch <N.Schurch at dundee.ac.uk>
>> To: bioconductor at r-project.org
>> Subject: [BioC] Problems with edgeR for differential expression.
>>
>> Dear Bioconductors,
>>
>> I'm trying to use edgeR to compute the differential expression from some
>> RNA-Seq data, but it seems to be generating some odd results.
>>
>> I'm running edgeR on data on about 6000 genes, across 5 experimental
>> conditions. When I compute the differential expression for any two of these
>> conditions edgeR it is returning a 'nan' fold-change for about 1500 out of
>> the 6000 genes being tested. Amazingly, it is also returning p-values for
>> these fold-changes, and the p-values cover a range of values from total
>> insignificant (>0.5) to very significant (<1E-4)! At first I thought if
>> might be because there is sometimes no signal for a gene in a given
>> condition, but 1) other cases of this nature produce '-inf' or 'inf'
>> fold-changes, not 'nan' fold-hanges, and 2) in some cases edgeR is
>> calculating a 'nan' fold-change for something that has signal in every
>> replicate of both conditions! I've checked everything I can think of... I
>> don't get any errors or warnings, the Norm Factors are sensible, the raw
>> signal is sensible, all the objects look well-formed (i.e. like they contain
>> all the bits they should contain).... its very confusing and frustrating.
>>
>> So, am I doing something wrong? Or is there a deeper problem with this
>> package?
>>
>> I'm using R v 2.13.1, edgeR 2.2.5 and the commands I'm using are:
>>
>>  # create the design matrix
>>> model.groups<-groups
>>> model.factors<-as.factor(**model.groups)
>>> model<-model.matrix(~model.**factors)
>>>
>>> # build DGElist and calculate normalization factors
>>> x=as.data.frame(data)
>>> rownames(x)=genenames
>>> y=DGEList(x,group=groups)
>>> y=calcNormFactors(y)
>>> str(y)
>>>
>>
>> Formal class 'DGEList' [package "edgeR"] with 1 slots
>>  ..@ .Data:List of 3
>>  .. ..$ :'data.frame': 17 obs. of  3 variables:
>>  .. .. ..$ group       : Factor w/ 5 levels "cond00","cond06",..: 3 3 3 4
>> 4
>> 4 3 3 1 1 ...
>>  .. .. ..$ lib.size    : num [1:17] 1866963 2364994 1712838 2782920
>> 2780054
>> ...
>>  .. .. ..$ norm.factors: num [1:17] 0.992 0.978 1.002 1.036 0.972 ...
>>  .. ..$ : num [1:6351, 1:17] 7 187 44 0 98 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:6351] "38536" "38537" "38538" "38539" ...
>>  .. .. .. ..$ : chr [1:17] "cond07.rep0020" "cond07.rep0021"
>> "cond07.rep0022" "cond08.rep0023" ...
>>  .. ..$ : Named logi [1:6351] FALSE FALSE FALSE FALSE FALSE FALSE ...
>>  .. .. ..- attr(*, "names")= chr [1:6351] "38536" "38537" "38538" "38539"
>> ...
>>
>>
>>> # estimate dispersion and fit models
>>> z=estimateGLMCommonDisp(y, design))
>>> str(z)
>>>
>>
>> Formal class 'DGEList' [package "edgeR"] with 1 slots
>>  ..@ .Data:List of 4
>>  .. ..$ :'data.frame': 17 obs. of  3 variables:
>>  .. .. ..$ group       : Factor w/ 5 levels "cond00","cond06",..: 3 3 3 4
>> 4
>> 4 3 3 1 1 ...
>>  .. .. ..$ lib.size    : num [1:17] 1866963 2364994 1712838 2782920
>> 2780054
>> ...
>>  .. .. ..$ norm.factors: num [1:17] 0.992 0.978 1.002 1.036 0.972 ...
>>  .. ..$ : num [1:6351, 1:17] 7 187 44 0 98 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:6351] "38536" "38537" "38538" "38539" ...
>>  .. .. .. ..$ : chr [1:17] "cond07.rep0020" "cond07.rep0021"
>> "cond07.rep0022" "cond08.rep0023" ...
>>  .. ..$ : Named logi [1:6351] FALSE FALSE FALSE FALSE FALSE FALSE ...
>>  .. .. ..- attr(*, "names")= chr [1:6351] "38536" "38537" "38538" "38539"
>> ...
>>  .. ..$ : num 0.101
>>
>>
>>> fit<-glmFit(z,model,**dispersion=z$common.**dispersion)
>>> str(fit)
>>>
>>
>> Formal class 'DGEGLM' [package "edgeR"] with 1 slots
>>  ..@ .Data:List of 12
>>  .. ..$ : num [1:6351, 1:5] -12.89 -9.22 -10.72 -11.54 -9.56 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:6351] "38536" "38537" "38538" "38539" ...
>>  .. .. .. ..$ : chr [1:5] "(Intercept)" "model.factorscond06"
>> "model.factorscond07" "model.factorscond08" ...
>>  .. ..$ : int [1:6351] 12 12 12 12 12 12 12 12 12 12 ...
>>  .. ..$ : Named num [1:6351] 23.3 3.45 30.57 25.89 3.07 ...
>>  .. .. ..- attr(*, "names")= chr [1:6351] "38536" "38537" "38538" "38539"
>> ...
>>  .. ..$ : num [1:17, 1:5] 1 1 1 1 1 1 1 1 1 1 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:17] "1" "2" "3" "4" ...
>>  .. .. .. ..$ : chr [1:5] "(Intercept)" "model.factorscond06"
>> "model.factorscond07" "model.factorscond08" ...
>>  .. .. ..- attr(*, "assign")= int [1:5] 0 1 1 1 1
>>  .. .. ..- attr(*, "contrasts")=List of 1
>>  .. .. .. ..$ model.factors: chr "contr.treatment"
>>  .. ..$ : num [1:6351, 1:17] 14.4 14.4 14.4 14.4 14.4 ...
>>  .. ..$ :'data.frame': 17 obs. of  3 variables:
>>  .. .. ..$ group       : Factor w/ 5 levels "cond00","cond06",..: 3 3 3 4
>> 4
>> 4 3 3 1 1 ...
>>  .. .. ..$ lib.size    : num [1:17] 1866963 2364994 1712838 2782920
>> 2780054
>> ...
>>  .. .. ..$ norm.factors: num [1:17] 0.992 0.978 1.002 1.036 0.972 ...
>>  .. ..$ : NULL
>>  .. ..$ : num 0.101
>>  .. ..$ : num [1:17] 1851365 2313681 1716797 2882487 2703330 ...
>>  .. ..$ : NULL
>>  .. ..$ : num [1:6351, 1:17] 6.32 160.53 59.13 12.7 105.68 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:6351] "38536" "38537" "38538" "38539" ...
>>  .. .. .. ..$ : chr [1:17] "cond07.rep0020" "cond07.rep0021"
>> "cond07.rep0022" "cond08.rep0023" ...
>>  .. ..$ : Named num [1:6351] -12.67 -9.33 -10.36 -11.66 -9.64 ...
>>  .. .. ..- attr(*, "names")= chr [1:6351] "38536" "38537" "38538" "38539"
>> ...
>>
>>
>>> # liklihood ratio statistics
>>> results=glmLRT(z, fit, coef = 4)
>>> str(results)
>>>
>>
>> Formal class 'DGELRT' [package "edgeR"] with 1 slots
>>  ..@ .Data:List of 10
>>  .. ..$ :'data.frame': 17 obs. of  3 variables:
>>  .. .. ..$ group       : Factor w/ 5 levels "cond00","cond06",..: 3 3 3 4
>> 4
>> 4 3 3 1 1 ...
>>  .. .. ..$ lib.size    : num [1:17] 1866963 2364994 1712838 2782920
>> 2780054
>> ...
>>  .. .. ..$ norm.factors: num [1:17] 0.992 0.978 1.002 1.036 0.972 ...
>>  .. ..$ : Named logi [1:6351] FALSE FALSE FALSE FALSE FALSE FALSE ...
>>  .. .. ..- attr(*, "names")= chr [1:6351] "38536" "38537" "38538" "38539"
>> ...
>>  .. ..$ : num 0.101
>>  .. ..$ :'data.frame': 6351 obs. of  4 variables:
>>  .. .. ..$ logConc     : num [1:6351] -12.67 -9.33 -10.36 -11.66 -9.64 ...
>>  .. .. ..$ logFC       : num [1:6351] 0.7057 -0.0157 0.5583 -0.3311
>> -0.1913
>> ...
>>  .. .. ..$ LR.statistic: num [1:6351] 1.38671 0.00165 1.83369 0.46732
>> 0.24042 ...
>>  .. .. ..$ p.value     : num [1:6351] 0.239 0.968 0.176 0.494 0.624 ...
>>  .. ..$ : num [1:6351, 1:5] -12.89 -9.22 -10.72 -11.54 -9.56 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:6351] "38536" "38537" "38538" "38539" ...
>>  .. .. .. ..$ : chr [1:5] "(Intercept)" "model.factorscond06"
>> "model.factorscond07" "model.factorscond08" ...
>>  .. ..$ : num [1:6351, 1:4] -12.63 -9.22 -10.52 -11.63 -9.62 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:6351] "38536" "38537" "38538" "38539" ...
>>  .. .. .. ..$ : chr [1:4] "(Intercept)" "model.factorscond07"
>> "model.factorscond08" "model.factorscond10"
>>  .. ..$ : num [1:17, 1:5] 1 1 1 1 1 1 1 1 1 1 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:17] "1" "2" "3" "4" ...
>>  .. .. .. ..$ : chr [1:5] "(Intercept)" "model.factorscond06"
>> "model.factorscond07" "model.factorscond08" ...
>>  .. .. ..- attr(*, "assign")= int [1:5] 0 1 1 1 1
>>  .. .. ..- attr(*, "contrasts")=List of 1
>>  .. .. .. ..$ model.factors: chr "contr.treatment"
>>  .. ..$ : num [1:17, 1:4] 1 1 1 1 1 1 1 1 1 1 ...
>>  .. .. ..- attr(*, "dimnames")=List of 2
>>  .. .. .. ..$ : chr [1:17] "1" "2" "3" "4" ...
>>  .. .. .. ..$ : chr [1:4] "(Intercept)" "model.factorscond07"
>> "model.factorscond08" "model.factorscond10"
>>  .. ..$ : num 0.101
>>  .. ..$ : chr "model.factorscond06"
>>
>> These are some examples of the results:
>>
>> --------------------
>> edgeR out:
>>
>> LR.statistic:  0.975075487675
>> logConc:  -13.3931239073
>> p.value:  0.323417620577
>> logFC:  NaN
>> geneID:  38956
>>
>> Raw data:
>>
>> Condition 1:
>> replicate 1: 2,372,532 good reads, female: 8.0
>> replicate 2: 3,966,968 good reads, female: no signal
>> replicate 3: 1,389,571 good reads, male: no signal
>>
>> Condition 2:
>> replicate 1: 3,102,608 good reads, male: 5.0
>> replicate 2: 3,451,983 good reads, male: no signal
>> replicate 3: 2,892,192 good reads, male: no signal
>> replicate 4: 3,620,124 good reads, male: 5.0
>> replicate 5: 2,640,968 good reads, female: no signal
>> --------------------
>>
>> --------------------
>> edgeR out:
>>
>> LR.statistic:  3.57045101322
>> logConc:  -13.6684814046
>> p.value:  0.0588163345523
>> logFC:  NaN
>> geneID:  38959
>>
>> Raw data:
>>
>> Condition 1:
>> replicate 1: 2,372,532 good reads, female: 5.0
>> replicate 2: 3,966,968 good reads, female: 5.0
>> replicate 3: 1,389,571 good reads, male: no signal
>>
>> Condition 2:
>>
>> replicate 1: 3,102,608 good reads, male: no signal
>> replicate 2: 3,451,983 good reads, male: no signal
>> replicate 3: 2,892,192 good reads, male: 7.0
>> replicate 4: 3,620,124 good reads, male: no signal
>> replicate 5: 2,640,968 good reads, female: no signal
>> --------------------
>>
>> --------------------
>> edgeR out:
>>
>>
>> LR.statistic:  1.81602638091
>> logConc:  -11.265720531
>> p.value:  0.177786996458
>> logFC:  NaN
>> geneID:  38965
>>
>> Raw data:
>>
>> Condition 1:
>>
>> replicate 1: 2,372,532 good reads, female: 22.0
>> replicate 2: 3,966,968 good reads, female: 27.0
>> replicate 3: 1,389,571 good reads, male: 5.0
>>
>> Condition 2:
>>
>> replicate 1: 3,102,608 good reads, male: 26.0
>> replicate 2: 3,451,983 good reads, male: 26.0
>> replicate 3: 2,892,192 good reads, male: 9.0
>> replicate 4: 3,620,124 good reads, male: 36.0
>> replicate 5: 2,640,968 good reads, female: 55.0
>> --------------------
>>
>>
>> --
>> Cheers,
>>
>> Nick Schurch
>>
>> Data Analysis Group (The Barton Group),
>> School of Life Sciences,
>> University of Dundee,
>> Dow St,
>> Dundee,
>> DD1 5EH,
>> Scotland,
>> UK
>>
>> Tel: +44 1382 388707
>> Fax: +44 1382 345 893
>>
>
> ______________________________**______________________________**__________
> The information in this email is confidential and intended solely for the
> addressee.
> You must not disclose, forward, print or use it without the permission of
> the sender.
> ______________________________**______________________________**__________
>
> --
> Cheers,
>
> Nick Schurch
>
> Data Analysis Group (The Barton Group),
> School of Life Sciences,
> University of Dundee,
> Dow St,
> Dundee,
> DD1 5EH,
> Scotland,
> UK
>
> Tel: +44 1382 388707
> Fax: +44 1382 345 893
>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: edgeRcomp_fullscale.png
Type: image/png
Size: 45012 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/bioconductor/attachments/20111011/478e7487/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: edgeRcomp_zoom.png
Type: image/png
Size: 69992 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/bioconductor/attachments/20111011/478e7487/attachment-0001.png>


More information about the Bioconductor mailing list