[BioC] 2x2 factorial loop without common reference (pool)
francois fauteux
francois.fauteux at gmail.com
Sun Apr 23 22:52:44 CEST 2006
Hi again;
Trying to work this out in the "two channels" way (I don't get clearly
what are the benefits of treating a two-color array as if it was two
one-color arrays)... If really necessary and clearly justified, I
might whant to switch to this type of analysis, but otherwise would
like to stay in what's more 'standard'.
Other try, but new design get error messages when lmfit, coefficients
not estimable 1 out of 2:
> fit4 <- lmFit(MAq, weights=w, design4)
Coefficients not estimable: b.a c.a d.c
> design4
a.b b.a a.c c.a b.d d.b c.d d.c
[1,] 1 -1 0 0 0 0 0 0
[2,] 1 -1 0 0 0 0 0 0
[3,] 1 -1 0 0 0 0 0 0
[4,] -1 1 0 0 0 0 0 0
[5,] -1 1 0 0 0 0 0 0
[6,] -1 1 0 0 0 0 0 0
[7,] 0 0 1 -1 0 0 0 0
[8,] 0 0 1 -1 0 0 0 0
[9,] 0 0 1 -1 0 0 0 0
[10,] 0 0 -1 1 0 0 0 0
[11,] 0 0 -1 1 0 0 0 0
[12,] 0 0 -1 1 0 0 0 0
[13,] 0 0 0 0 0 -1 0 0
[14,] 0 0 0 0 1 -1 0 0
[15,] 0 0 0 0 1 -1 0 0
[16,] 0 0 0 0 1 1 0 0
[17,] 0 0 0 0 -1 1 0 0
[18,] 0 0 0 0 -1 1 0 0
[19,] 0 0 0 0 -1 0 1 -1
[20,] 0 0 0 0 0 0 1 -1
[21,] 0 0 0 0 0 0 1 -1
[22,] 0 0 0 0 0 0 -1 1
[23,] 0 0 0 0 0 0 -1 1
[24,] 0 0 0 0 0 0 -1 1
Why are Coefficients not estimable??? Thought this design would take
into account the dye effects...
Second design try:
design3 <- modelMatrix(targets,ref="a")
> contrast.matrix <- makeContrasts(b,c,d-b,d-c,
+ levels=design3)
> contrast.matrix
b c d - b d - c
c 0 1 0 -1
b 1 0 -1 0
d 0 0 1 1
This seems correct a priori. If you could confirm that it is OK, and
that this is indeed the best way of working things out...
Many thanks, regards.
François
On 4/23/06, Naomi Altman <naomi at stat.psu.edu> wrote:
> I would use single channel analysis for
> this. The only problem is that Limma allows only
> 1 level of random effects. Hence, you will need to average the dye-swaps.
>
> Anyways with single channel analysis, you have a
> balanced incomplete block design with factorial
> treatments, and the analysis is much simplified.
>
> --Naomi
>
> At 01:41 PM 4/23/2006, francois fauteux wrote:
> >Hi;
> >
> >We are doing an experiment with agilent 44K (3 biological reps,
> >complete dye-swap):
> >
> >a - control
> >b - treatment 1
> >c - treatment 2
> >d - treatment 1 + treatment 2
> >
> >and I would like to output evidence of the interaction between two
> >treatments and effect on gene expression.
> >
> >24 chips:
> >
> >SlideNumber Cy3 Cy5
> >1 a1 b1
> >2 a2 b2
> >3 a3 b3
> >4 b1 a1
> >5 b2 a2
> >6 b3 a3
> >7 a1 c1
> >8 a2 c2
> >9 a3 c3
> >10 c1 a1
> >11 c2 a2
> >12 c3 a3
> >13 b1 d1
> >14 b2 d2
> >15 b3 d3
> >16 d1 b1
> >17 d2 b2
> >18 d3 b3
> >19 c1 d1
> >20 c2 d2
> >21 c3 d3
> >22 d1 c1
> >23 d2 c2
> >24 d3 c3
> >
> >I've done several tests with limma to isolate significant results in
> >the following:
> >1- a vs b;
> >2- a vs c;
> >3- b bs d;
> >4- c vs d;
> >
> >with this "targets.txt":
> >
> >SlideNumber Cy3 Cy5
> >1 a b
> >2 a b
> >3 a b
> >4 b a
> >5 b a
> >6 b a
> >7 a c
> >8 a c
> >9 a c
> >10 c a
> >11 c a
> >12 c a
> >13 b d
> >14 b d
> >15 b d
> >16 d b
> >17 d b
> >18 d b
> >19 c d
> >20 c d
> >21 c d
> >22 d c
> >23 d c
> >24 d c
> >
> >First option:
> >
> > > f <- paste(targets$Cy3, targets$Cy5, sep = ".")
> > > f <- factor(f, levels = c("a.b", "b.a",
> > "a.c", "c.a", "b.d", "d.a", "c.d", "d.a"))
> > > design1 <- model.matrix(~0 + f)
> >
> > > design
> > a.b b.a a.c c.a b.d d.b c.d d.c
> >1 1 0 0 0 0 0 0 0
> >2 1 0 0 0 0 0 0 0
> >3 1 0 0 0 0 0 0 0
> >4 0 1 0 0 0 0 0 0
> >5 0 1 0 0 0 0 0 0
> >6 0 1 0 0 0 0 0 0
> >7 0 0 1 0 0 0 0 0
> >8 0 0 1 0 0 0 0 0
> >9 0 0 1 0 0 0 0 0
> >10 0 0 0 1 0 0 0 0
> >11 0 0 0 1 0 0 0 0
> >12 0 0 0 1 0 0 0 0
> >13 0 0 0 0 1 0 0 0
> >14 0 0 0 0 1 0 0 0
> >15 0 0 0 0 1 0 0 0
> >16 0 0 0 0 0 1 0 0
> >17 0 0 0 0 0 1 0 0
> >18 0 0 0 0 0 1 0 0
> >19 0 0 0 0 0 0 1 0
> >20 0 0 0 0 0 0 1 0
> >21 0 0 0 0 0 0 1 0
> >22 0 0 0 0 0 0 0 1
> >23 0 0 0 0 0 0 0 1
> >24 0 0 0 0 0 0 0 1
> >
> >This gives significant results for each one of the "levels" but does
> >not take into account the dye-swap (i.e "a.b" and "b.a" are considered
> >independent).
> >
> >Other tested option is:
> > > design2 <- modelMatrix(targets,ref="a")
> >
> > > design
> > p s sp
> >ab1 0 1 0
> >ab2 0 1 0
> >ab3 0 1 0
> >ba1 0 -1 0
> >ba2 0 -1 0
> >ba3 0 -1 0
> >ac1 1 0 0
> >ac2 1 0 0
> >ac3 1 0 0
> >ca1 -1 0 0
> >ca2 -1 0 0
> >ca3 -1 0 0
> >bd1 0 -1 1
> >bd2 0 -1 1
> >bd3 0 -1 1
> >db1 0 1 -1
> >db2 0 1 -1
> >db3 0 1 -1
> >cd1 -1 0 1
> >cd2 -1 0 1
> >cd3 -1 0 1
> >dc1 1 0 -1
> >dc2 1 0 -1
> >dc3 1 0 -1
> >
> >This gives results for "b" effect, "c" effect, and "d" effect.
> >However, I could'nt get results for the 4 comparisons of interest
> >(even though the matrix is coherent).
> >
> >Questions:
> >
> >1 - What would be the best option (design and operations) to get to
> >contrasts of interest considering that the experiment has a 4
> >treatments in a factorial design without common reference (a vs b, a
> >vs c, b vs d, c vs d) and taking into account the dye-effect;
> >
> >2- Is this method (4 contrasts) the best one considering that
> >treatment "d" is a combination of treatments "b" and "c" (factorial
> >type design). How could one directly get to identify genes
> >differentially expressed due to the interaction between treatment "b"
> >and treatment "c" (i.e effect of "d" over "b" and "c").
> >
> >In Limma Users Guide and elsewhere on this forum, I could not find a
> >clear description of how this type of analysis should be performed,
> >even though it is a simple design (i.e 2X2 factorial without a common
> >reference - two color arrays - complete dye swap).
> >
> >Thanks for your time, best regards.
> >
> >François Fauteux
> >Étudiant à la maîtrise en biologie végétale
> >Centre de recherche en horticulture
> >Université Laval
> >francois.fauteux at gmail.com
> >
> >_______________________________________________
> >Bioconductor mailing list
> >Bioconductor at stat.math.ethz.ch
> >https://stat.ethz.ch/mailman/listinfo/bioconductor
> >Search the archives:
> >http://news.gmane.org/gmane.science.biology.informatics.conductor
>
> Naomi S. Altman 814-865-3791 (voice)
> Associate Professor
> Dept. of Statistics 814-863-7114 (fax)
> Penn State University 814-865-1348 (Statistics)
> University Park, PA 16802-2111
>
>
More information about the Bioconductor
mailing list