[BioC] design matrix question

Marcelo Luiz de Laia mlaia at fcav.unesp.br
Fri Feb 3 04:07:21 CET 2006


Hello,

After looking at the users guide, I have begun to analyse our data.

I found problems for to do a proper design matrix, with my biologist 
background.

We are trying to analyse a 4x2x3 factorial design. We have 3 biological 
replicates.

I made a search in the list archives and found a lot off messages about 
design in limma, but I steel with doubts about it.

I would like to understand the design matrix.

For example, I have this targets in a pData (made by hand)
            
FileName Variedade Treatment Time
1    File01      Var1       Con   T1
2    File02      Var1       Tra   T1
3    File03      Var1       Con   T1
4    File04      Var1       Tra   T1
5    File05      Var1       Con   T1
6    File06      Var1       Tra   T1
7    File07      Var1       Con   T2
8    File08      Var1       Tra   T2
9    File09      Var1       Con   T2
10   File10      Var1       Tra   T2
11   File11      Var1       Con   T2
12   File12      Var1       Tra   T2
13   File13      Var1       Con   T3
14   File14      Var1       Tra   T3
15   File15      Var1       Con   T3
16   File16      Var1       Tra   T3
17   File17      Var1       Con   T3
18   File18      Var1       Tra   T3
(...) 
72   File 72     Var4       Tra   T3

Then, I  made this design matrix, by hand:

design <-
  model.matrix(~-1+factor(c(1,2,1,2,1,2,3,4,3,4,3,4,5,6,5,6,5,6,
                            7,8,7,8,7,8,9,10,9,10,9,10,11,12,11,12,11,12,
                            13,14,13,14,13,14,15,16,15,16,15,16,
                            17,18,17,18,17,18,19,20,19,20,19,20,
                            21,22,21,22,21,22,23,24,23,24,23,24)))
colnames(design) <- c("Ctrl_v1a1","Data_v1a1","Ctrl_v1a2","Data_v1a2",
                      "Ctrl_v1a3","Data_v1a3",
                      "Ctrl_v2a1","Data_v2a1","Ctrl_v2a2","Data_v2a2",
                      "Ctrl_v2a3","Data_v2a3",
                      "Ctrl_v3a1","Data_v3a1","Ctrl_v3a2","Data_v3a2",
                      "Ctrl_v3a3","Data_v3a3",
                      "Ctrl_v4a1","Data_v4a1","Ctrl_v4a2","Data_v4a2",
                      "Ctrl_v4a3","Data_v4a3")

I would like to know what is in the topTable function coef=1? With this 
answer I will be able to understand what are in the others coef.

At this step am I impossible to get differently expressed genes for Var 
1 on time 1? Tra is treated and Con is a control (not treated).

After this step, I made a contrast matrix in this way:

contrast.matrix <-
  makeContrasts("Data_v1a1-Ctrl_v1a1","Data_v1a2-Ctrl_v1a2",
                          "Data_v1a3-Ctrl_v1a3",
                           "Data_v2a1-Ctrl_v2a1","Data_v2a2-Ctrl_v2a2",
                          "Data_v2a3-Ctrl_v2a3",
                         "Data_v3a1-Ctrl_v3a1","Data_v3a2-Ctrl_v3a2",
                          "Data_v3a3-Ctrl_v3a3",
                         "Data_v4a1-Ctrl_v4a1","Data_v4a2-Ctrl_v4a2",
                          "Data_v4a3-Ctrl_v4a3",
                            levels=design)

In this step I have the differentially expressed genes for Var 1 on time 
1 in the coef=1 in topTable function?

With this design we are able to get the differentially expressed genes 
in time 3 vs time 2? And for treated vs control?

I try to do another design based on targets:

design<-model.matrix(~Variedade*Treatment*Time, data=pData(targets))

In this case, I have a intercept and more 23 columns.

Is this case, what comparison are in the topTable functions coef=1? And 
in coef=2? And in coef=23?

Need I to do a contrast matrix, too?

In this case I not need to makecontrasts for get the expressed genes for 
Var 4 vs Var 1 in all times, i.e., considering the treated and control 
included?

I guess if I solve these doubts I will made my design matrix correct.

So, I decide to ask for some advice from Bioconductor´s list.

I am interested in several aspects (contrasts), like differences between 
Variedade types with treatment, and differences between Variedade types 
without treatment (time excluded or included).

Which contrasts can answer these questions?

Thanks in advance for your assistance.

-- 
Marcelo Luiz de Laia
Ph.D Candidate
São Paulo State University (http://www.unesp.br/eng/)
School of Agricultural and Veterinary Sciences
Department of  Technology
Via de Acesso Prof.Paulo Donato Castellane s/n
14884-900   Jaboticabal - SP - Brazil
Phone: +55-016-3209-2675
Cell: +55-016-97098526



More information about the Bioconductor mailing list