Question: RES: Genechip experiment in limma
0
gravatar for Gordon Smyth
15.9 years ago by
Gordon Smyth37k
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Gordon Smyth37k wrote:
> My spatial heterogeneity is intra-print-tip. The unit must be smaller > that the sector. > > Your assumption about my Genechip hybridization classification is > correct. > (C01h, C02h, C04h, C06h, A01h, A02h, A04h, A06h, B01h, B02h, B04h, > B06h). > > I want to know what is differentially expressed between Trt A x Ctrl, > Trt B x Ctrl, Trt x Ctrl and Trt A x Trt B. > >>Yes, there is a design matrix which incorporates the temporal > information, >>but it is difficult to do statistical analysis in the absence of any >> replication. (I do wish people would be prepared to invest in an extra > two >>or three chips in order to be able to do statistical analysis.) > > The absence of replication is not caused by money but by lack of > biological material. > I really want to evaluate the DE with a homogeneous pool of starting > cells. In this case, there was material for just 12 hybridizations. In > my sense, when doing time series experiments, and you have a limited > amount of material, it is better to increase the number of time points > instead of the number of replicates by time points. And the temporal > information can be used as "replication" (the 4 time points for each > sample are from the same cell pool, but they have different fates since > the beggining of the experiment. I do not take aliquots from the same > cell culture at 1h, 2h, 4h and 6h). I know they are not biological > replicates in a strict sense, but my study objectives and implications > are "aware" of this. > > So, my question is: > > May I create a design and contrast matrix that use the time points as > replicates, but in a more flexible way. For instance, taking the first > three time points as replicates of a early process, and the last three > time points as replicates of a late process, without doing three > diferent analysis? Well, you would have to define two different design matrices and do two separate fits. Gordon > Christian > > -----Mensagem original----- > De: bioconductor-bounces@stat.math.ethz.ch > [mailto:bioconductor-bounces@stat.math.ethz.ch] Em nome de Gordon Smyth > Enviada em: sexta-feira, 22 de agosto de 2003 03:05 > Para: Christian Probst > Cc: BioC Mailing List > Assunto: [BioC] Genechip experiment in limma > > At 02:48 AM 22/08/2003, Christian Probst wrote: >>Gordon, >> >>I succeed in loading the cDNA microarray data. As I need a spatial >> normalization, I will try to do it before the use of limma package. > > Remember that print-tip-loess is in itself a simple form of spatial > normalization (as well as intensity-based normalization). In my > experience, > print-tip-loess does most of what spatial normalization would have done. > > But, as you say, if you really do want to do formal 2D normalization > you'll > have to do it outside of limma. > >>I am also running a Genechip experiment, with four time points and > three >>treatments (A,B and Ctrl), with no replicates. > > Am I correct in assuming that your first 4 chips are hybridised with > Ctrl, > the second four with A and the last four with B? Are the time points > 1,2,3,4,1,2,3,4,1,2,3,4? > >>I would like to ask you about the design table in order to analyse this >> data. >>I want to find those genes that are differentially expressed in AxCtrl, >> BxCtrl, and AB x Ctrl. > > Do you want genes which are DE between A and Ctrl and DE between B and > Ctrl? Obviously this includes identifying genes which are both DE for A > vs > Ctrl and B vs Ctrl. I am not sure what you mean by "AB x Ctrl", perhaps > you > just mean different for both A vs Ctrl and B vs Ctrl. > >>Is this design matrix correct? > > No it's not. If my assumptions above are correct, and your time points > were > really replicates, which they're not, then you'd want either > > > design > Ctrl A B > 1 0 0 > 1 0 0 > 1 0 0 > 1 0 0 > 0 1 0 > 0 1 0 > 0 1 0 > 0 1 0 > 0 0 1 > 0 0 1 > 0 0 1 > 0 0 1 > > with contrast matrix > > > cont.matrix > A-Ctrl B-Ctrl > -1 -1 > 1 0 > 0 1 > > > design > Ctrl A-Ctrl B-Ctrl > 1 0 0 > 1 0 0 > 1 0 0 > 1 0 0 > 1 1 0 > 1 1 0 > 1 1 0 > 1 1 0 > 1 0 1 > 1 0 1 > 1 0 1 > 1 0 1 > > with contrast matrix > > > cont.matrix > A-Ctrl B-Ctrl > 0 0 > 1 0 > 0 1 > > I suggest using the first design matrix, in which case using limma > 1.1.11, > you would go > > fit <- lmFit(eset, design) > fit <- contrasts.fit(fit, cont.matrix) > fit <- eBayes(fit) > clas <- classifyTests(fit) > vennDiagram(clas) > > to classify each gene as differentially expressed or not for AvsCtrl, > BvsCtrl or both. > >> >design >> A B AB >>1 0 0 0 >>2 0 0 0 >>3 0 0 0 >>4 0 0 0 >>5 1 0 1 >>6 1 0 1 >>7 1 0 1 >>8 1 0 1 >>9 0 1 1 >>10 0 1 1 >>11 0 1 1 >>12 0 1 1 >> >>As I have a time-course experiment, and I am using the four time points >> as replicates, in order to analyze this data statistically, I wonder if >> there is a design table that uses the temporal information in a > flexible >>way, to identify DEG in the beggining or ending of the infection >>proccess (I will mix my "replication" with my "temporal framework"). > > Yes, there is a design matrix which incorporates the temporal > information, > but it is difficult to do statistical analysis in the absence of any > replication. (I do wish people would be prepared to invest in an extra > two > or three chips in order to be able to do statistical analysis.) > > There are things you could do, for example using genes which are not > differentially expressed to estimate variability for genes which are > differentially expressed, but this is a research problem. > > Regards > Gordon > >>TIA >> >>Christian >> >> >>-----Mensagem original----- >>De: Gordon Smyth [mailto:smyth@wehi.edu.au] >>Enviada em: ter?a-feira, 19 de agosto de 2003 23:01 >>Para: Christian M. Probst >>Assunto: Re: Limma error >> >>Christian, >> >>At 02:50 AM 20/08/2003, Christian M. Probst wrote: >> >Dear Mr. Smyth, >> > >> >I am trying to use the limma package, but in the first step of the > data >> >loading, I run into the following error: >> > >> > > library(limma) >> > > files<-dir(pattern="*\\.spot") >> > > files >> >[1] "EP3_05_ET5_05_0808.spot" "EP3_28_ET5_28_0908.spot" >> >[3] "SP3_30_EP5_28_0908.spot" "SP3_30_ST5_30_0908.spot" >> > > RG<-read.maimages(files,source="spot") >> >Read EP3_05_ET5_05_0808.spot >> >Read EP3_28_ET5_28_0908.spot >> >Error in "[<-"(*tmp*, , i, value = NULL) : >> > number of items to replace is not a multiple of replacement >>length >> > > >> > >> >This message appears consistently in other files, also. I havent > found >>a >> >description in the Bioconductor discussion list, so I choose to > bother >>you >> >directly. >> >>It is pretty hard to give you a lot of help from the above information. >> It >>is not true to say that read.maimages consistently gives this error. > The >> >>above output shows that read.maimages read your first two files >>"EP3_05_ET5_05_0808.spot" and "EP3_28_ET5_28_0908.spot" successfully > but >> >>then fails on "SP3_30_EP5_28_0908.spot". What is different about this >> third >>file compared the first two? >> >>Also, what version of limma are you using? >> >>Gordon >> >> >> >> >TIA >> > >> >Christian > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor > > Esta mensagem foi verificada pelo E-mail Protegido Terra. > Scan engine: VirusScan / Atualizado em 20/08/2003 / Vers?o: 1.3.13 > Proteja o seu e-mail Terra: http://www.emailprotegido.terra.com.br/
ADD COMMENTlink written 15.9 years ago by Gordon Smyth37k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 183 users visited in the last hour