Question: Pair-wise Limma Analysis with time series data and two levels (control and treatment)
0
gravatar for kawess
10 weeks ago by
kawess0
Canada
kawess0 wrote:

Hello, I am a computer scientist who is new at analyzing microarray data. So my question may be too easy so I apologize for that. I really need help with that. I got some dataset on GEO (GSE44248) on the human LCL cell line. I want to analyze this data using Limma to find differentially expressed genes. My data are time series and I want to find differentially expressed genes over time. Here is the description of the data:

  • 8 LCLs samples (Individual in the picture)
  • Time points: 2, 4, 8, 12, 16 and 24 hrs -Treatment: Dex vs EtOH (control) -Each sample is either treated with Dex or EtOH so I am in a case of paired-sample.

Here is a snapshot of my data matrix : enter image description here

My questions are the following:

How can get a good design matrix for Limma to analyze my paired sample time series data ?

I was thinking about this I will use spline for time: b=ns(Timepoint,df=3) : ~ Time_point*Treatment+ Individual

Do you think is a good idea? If not which formula do advice to use if I want to find differentially expressed genes between treated and control sample for at least one time point and across the cell line samples?

Which coefficient should I check in the topTable function ?

Thank you very much in advance.

ADD COMMENTlink modified 10 weeks ago by Gordon Smyth39k • written 10 weeks ago by kawess0
Answer: Pair-wise Limma Analysis with time series data and two levels (control and trea
1
gravatar for Gordon Smyth
10 weeks ago by
Gordon Smyth39k
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Gordon Smyth39k wrote:

The dataset GSE44248 uses HT-12 Illumina Beadchips, so I would recommend using neqc in the limma package to do the normalizing and preprocessing. See the case study in Section 17.3 of the limma User Guide.

As far as a design matrix is concerned, I can't see why you would not use the standard approach that we recommend for most experiments, for example in Section 9.6.1 of the limma User Guide. Just combine time and treatment into a single factor:

Group <- factor(paste(Time_point, Treatment, sep="."))

create the design matrix by

design <- model.matrix(~ 0 + Group + Individual)
colnames(design)[1:12] <- levels(Group)

and form contrasts in the usual way to test any hypothesis you want.

ADD COMMENTlink modified 10 weeks ago • written 10 weeks ago by Gordon Smyth39k

Thank you Gordon . My problem was because of the blocks. I will read the examples in the user manual. Thank you again

ADD REPLYlink written 10 weeks ago by kawess0
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: 154 users visited in the last hour