Dear all, I am writing in relation to some doubts that I have on the implementation of a design matrix in a factorial design experiment. I have tried to follow, religiously, the recommendations as at paragraph 23.10 of the R&Bioconductor monography, but have come up with several afterthoughts. Here is my experiment. Together with my colleagues, we have taken a human cell culture, the human lymphoblastoid line TK6, and split it in two laboratories with different environmental conditions. Specifically, environmental background ionizing radiation differs across the labs by a factor of 80. We are asking whether continuous cell culture under reduced environmental background radiation renders cells more or less tolerant to acute exposure to genotoxic or cytotoxic agents, a line of research that the lab in which I am working has been busy for several years. These two labs, and here is the first factor, are called "ISS" and "LNGS" in what follows. We have isolated total RNA samples (all instances contain 4 biological replicates) from these two cell cultures after they have been maintained in each laboratory for six months (corresponding to hundreds of cell generations) since we hypothesized that it took several cell generations before deviations at the level of the transcriptome may reveal. Therefore, a second variable (factor) is time, and this has the values "0" and "6 months". As you imagine, we have a determination on RNA extracted from the "0" time sample, which is the 'mother' cell culture of both "ISS" and "LNGS" (6 months) cultures. Additionally, to reveal any subtle differences between the two aged cultures, we have exposed them to an acute dose of ionizing radiation, which acts as a stressor. Our third variable/factor, therefore, is radiation Dose, which takes the values "0" and "1 Gy". Accordingly, we have an RNA extraction and a microarray determination at time = "0" and Dose = "1Gy", to compare with responses to radiation after six months in each laboratory. Following the BioC Monography, I have applied the model.matrix() function obtaining a design matrix with 8 columns. Two such columns, howewer, have no sense to this particular experiment, since the variable Laboratory ("ISS", "LNGS") has no meaning unless 6 months have elapsed, since their source is the same culture, the "0" culture from which the experiment started. Accordingly, I deleted - by hand - those two columns and obtained the attached design matrix. Do you think that this matrix is correct? We have applied it for some time in our analysis and felt sure it was, but we have come across to some odd results with respect to all analyses of 1Gy-irradiated samples and we are now re-evaluating this matrix. One anchoring point in our reasoning is what follows. The samples exposed to 1Gy after six months in culture in both labs (ISS, LNGS) are not directly derived from samples exposed to the same 1Gy dose at time "0". Those would be survivors to the 1Gy dose delivered 6 months before, which is not a question being addressed in this experiment (though a reasonable question). Instead, samples exposed to 1Gy are always derived from their unirradiated counterparts at the same time and in the same laboratory. For example, the LNGS, 6 months, 1Gy-irradiated sample is obtained from exposure to 1Gy of the sample maintained for six months at LNGS, again, not by maintaining for six months at LNGS the sample that had been exposed to 1Gy dose at time "0". I want to thank you very much in advance for the patience that you have put in reading all down here and I hope to read any suggestions. yours, Massimo P. -------------- next part -------------- > design (Intercept) Dose1Gy Ageing6mo Dose1Gy:Ageing6mo Ageing6mo:LabLNGS Dose1Gy:Ageing6mo:LabLNGS 1 1 0 0 0 0 0 2 1 0 0 0 0 0 3 1 0 0 0 0 0 4 1 0 0 0 0 0 5 1 1 0 0 0 0 6 1 1 0 0 0 0 7 1 1 0 0 0 0 8 1 1 0 0 0 0 9 1 0 1 0 0 0 10 1 0 1 0 0 0 11 1 0 1 0 0 0 12 1 0 1 0 0 0 13 1 1 1 1 0 0 14 1 1 1 1 0 0 15 1 1 1 1 0 0 16 1 1 1 1 0 0 17 1 0 1 0 1 0 18 1 0 1 0 1 0 19 1 0 1 0 1 0 20 1 0 1 0 1 0 21 1 1 1 1 1 1 22 1 1 1 1 1 1 23 1 1 1 1 1 1 24 1 1 1 1 1 1

I thought I should add here how the design matrix was created. Probably, if I understand the language correctly, this is a mixed factorial/nested design with three two-level factors: Dose <- factor(targets$Dose, levels=c("Cn", "1Gy")) # It's very important that you define the levels now, with their order, as otherwise he will do it for you and this may not be the way you want it. Lab <- factor(targets$Lab, levels=c("ISS", "LNGS")) Ageing <- factor(targets$Ageing, levels=c("t0", "6mo")) disegno.3 <- model.matrix(~Dose + Ageing + Dose:Ageing + Ageing:Lab + Dose:Ageing:Lab +1) Yours, Massimo Massimo Pinto Post Doctoral Research Fellow Enrico Fermi Centre and Italian Public Health Research Institute (ISS) Rome, Italy http://claimid.com/massimopinto On Thu, Feb 11, 2010 at 11:07, Massimo Pinto <pintarello at="" gmail.com=""> wrote: > Dear all, > > I am writing in relation to some doubts that I have on the > implementation of a design matrix in a factorial design experiment. I > have tried to follow, religiously, the recommendations as at paragraph > 23.10 of the R&Bioconductor monography, but have come up with several > afterthoughts. > > Here is my experiment. Together with my colleagues, we have taken a human cell > culture, the human lymphoblastoid line TK6, and split it in two > laboratories with > different environmental conditions. Specifically, environmental > background ionizing radiation differs across the labs by a factor of > 80. We are asking whether continuous cell culture under reduced > environmental background radiation renders cells more or less tolerant > to acute exposure to genotoxic or cytotoxic agents, a line of research > that the lab in which I am working has been busy for several years. > These two labs, and here is the first factor, are called "ISS" and > "LNGS" in what follows. We have > isolated total RNA samples (all instances contain 4 biological > replicates) from these two cell cultures after they have been > maintained in each laboratory for six months (corresponding to hundreds of cell > generations) since we hypothesized that it took several cell > generations before deviations at the level of the transcriptome may > reveal. Therefore, a second variable (factor) is time, and this has > the values "0" and "6 months". > As you imagine, we have a determination on RNA extracted from the "0" > time sample, which is the 'mother' cell > culture of both "ISS" and "LNGS" (6 months) cultures. > Additionally, to reveal any subtle differences between the two aged > cultures, we have exposed them to an acute dose of ionizing radiation, > which acts as a stressor. Our third variable/factor, therefore, is radiation > Dose, which takes the values "0" and "1 Gy". Accordingly, we have an > RNA extraction and a microarray determination at time = "0" and Dose = > "1Gy", to compare with responses ?to radiation after six months in > each laboratory. > Following the BioC Monography, I have applied the model.matrix() > function obtaining > a design matrix with 8 columns. Two such columns, howewer, have no > sense to this particular experiment, since the variable Laboratory > ("ISS", "LNGS") > has no meaning unless 6 months have elapsed, since their source is the > same culture, the "0" culture from which the experiment started. > Accordingly, I deleted - by hand - those two columns and obtained the > attached design matrix. > Do you think that this matrix is correct? We have applied it for some > time in our analysis and felt sure it was, but we have come across to > some odd results with respect to all analyses of 1Gy-irradiated > samples and we are now re-evaluating this matrix. > > One anchoring point in our reasoning is what follows. The samples > exposed to 1Gy after six months in culture in both labs (ISS, LNGS) > are not directly derived from samples exposed to the same 1Gy dose at > time "0". Those would be survivors to the 1Gy dose delivered 6 months > before, which is not a question being addressed in this experiment > (though a reasonable question). Instead, samples exposed to 1Gy are > always derived from > their unirradiated counterparts at the same time and in the same > laboratory. For example, the LNGS, 6 months, 1Gy-irradiated sample is obtained > from exposure to 1Gy of the sample maintained for six months at LNGS, > again, not by maintaining for six months at LNGS the sample that had been > exposed to 1Gy dose at time "0". > > I want to thank you very much in advance for the patience that you > have put in reading all down here and I hope to read any suggestions. > > yours, > > Massimo P. >