Hi,

In edgeR I performed LRT for all groups against the reference like
this:

> lrtAll = glmLRT(fit, coef=2:10)

I wonder if it is possible to use decideTestsDGE() to find out the
number of upregulated DE genes? I can use it when I perform LRT to
compare one group against the reference but on lrtAll I get the error
message:

> decideTestsDGE(lrtAll)
Error in array(x, c(length(x), 1L), if (!is.null(names(x))) list(names(x),  :
  'data' must be of a vector type, was 'NULL'

> sessionInfo()
R version 3.1.0 (2014-04-10)
Platform: x86_64-apple-darwin10.8.0 (64-bit)

locale:
[1] C

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

other attached packages:
[1] edgeR_3.6.4  limma_3.20.8

loaded via a namespace (and not attached):
 [1] AnnotationDbi_1.26.0    Biobase_2.24.0  BiocGenerics_0.10.0     DBI_0.2-7
 [5] DESeq2_1.4.5           GenomeInfoDb_1.0.2  GenomicRanges_1.16.3    IRanges_1.22.9
 [9] RColorBrewer_1.0-5      RSQLite_0.11.4         Rcpp_0.11.2  RcppArmadillo_0.4.320.0
[13] XML_3.98-1.1           XVector_0.4.0          annotate_1.42.0  genefilter_1.46.1
[17] geneplotter_1.42.0      grid_3.1.0             lattice_0.20-29  locfit_1.5-9.1
[21] parallel_3.1.0         stats4_3.1.0           survival_2.37-7  tools_3.1.0
[25] xtable_1.7-3

---------------------------------------------------------------- 
Jon Bråte
Section for Genetics and Evolutionary Biology (EVOGENE)
Department of Biosciences
University of Oslo
P.B. 1066 Blindern
N-0316, Norway
Email: jon.brate@ibv.uio.no
Phone: 922 44 582
Web: mn.uio.no/ibv/english/people/aca/jonbra/index.html

Dear Jon,

No, you can't use decideTestsDGE() to count the number of DE genes when doing a test for multiple contrasts at once. The whole purpose of decideTestDGE is to count up and down regulated genes. However when you are doing a test for multiple coefficients or contrasts, then the test is analogous to an F-test on more than one degree of freedom. There is no concept of directionality for an F-test, i.e., you can't classify a significant result as being up or down, just not all zero.

The same situation exists for the corresponding function in limma. decideTests() can be used for t-tests but not for F-tests.

Best wishes
Gordon