[BioC] question about design for limma time course, 2 conditions and drug treatment microarray experiment

Pekka Kohonen pkpekka at gmail.com
Mon Nov 25 17:35:16 CET 2013


Dear Ninni,

I have used limma/bioconductor for dose-response analysis but am not
going to answer your limma question directly (others can probably do
better than I). There is, however, another package called "IsoGeneGUI"
for doing dose-response analysis in a model-based manner. It is using
a CRAN package called Isogene. I have been reading a book on
dose-response analysis that describes the methodology in detail
(though the vignette gives enough information to do the analysis).
Essentially, model based dose-response analysis is a different
paradigm and, it seems to me, cannot be carried out using limma. Since
Isogene uses SAM statistics it is not as powerful / sensitive as limma
which uses variance shrinkage and parametric statistics (instead of
permutations). But it should be very reliable and reasonably sensitive
(though SAM-like statistics would especially benefit from having at
least 3 biological replicates).

I think that model-based dose response analysis makes sense but is a
bit of an undeveloped area, possibly because academic investigators
have not had so many doses to play around with in the past (and you
seem to have quite a few dose in your data). This is now changing when
omics profiling is becoming cheaper and there is a debate raging over
monotonic vs. non-monotonic dose responses in relation to the hormone
mimicking chemicals in the environment, for instance. The authors of
IsogeneGUI include people in the pharmaceutical industry also.

Best Regards, Pekka

2013/11/22 Ninni Nahm [guest] <guest at bioconductor.org>:
>
> Dear list,
>
> I have a conceptual question about creating a design matrix for a more complicated experimental design.
>
> I have microarray data of
> - two different conditions (treatment/control),
> - over a series of time points (20, 45, 90, 180 minutes)
> - and different dose concentrations of a certain drug (no treatment, 1mg, 2mg, 3mg, 4mg, 5mg).
> - and I have 2 replicates per drug, time point, and condition
> I think, I know how to do it when I want to consider time only (please correct me when I'm wrong!):
>
> ## find genes which change over time differently between the treatment and the control.
>
> cont.dif <- makeContrasts(
> Dif1  = (treatment_1mg_tp45-treatment_1mg_tp20)-(control_tp45-control_tp20),
> Dif2  = (treatment_1mg_tp90-treatment_1mg_tp45)-(control_tp90-control_tp45),
> Dif3  = (treatment_1mg_tp180-treatment_1mg_tp90)-(control_tp180-control_tp90),
> levels=design)
>
> However, I would like to know which genes are changing over time and drug exposure between control and treatment.
> Would I have to do the contrasts for every dose?
>
> Any help is much appreciated!
>
> Best,
> Ninni
>
>
>
>  -- output of sessionInfo():
>
>> sessionInfo()
> R version 3.0.2 (2013-09-25)
> Platform: x86_64-pc-linux-gnu (64-bit)
>
> locale:
>  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
>  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
>  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
>  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
>  [9] LC_ADDRESS=C               LC_TELEPHONE=C
> [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
>
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
>
> other attached packages:
> [1] sva_3.8.0     mgcv_1.7-27   nlme_3.1-113  corpcor_1.6.6 limma_3.18.3
>
> loaded via a namespace (and not attached):
> [1] grid_3.0.2      lattice_0.20-24 Matrix_1.1-0
>>
>
>
> --
> Sent via the guest posting facility at bioconductor.org.
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor



More information about the Bioconductor mailing list