Search
Question: intra-sample time series comparison using DESeq2
0
gravatar for prppmet
6 weeks ago by
prppmet0
prppmet0 wrote:

I need help with Time series testing using DESeq2.
- I have 5 subjects given same treatment at 11 different time points (for each subject: single observation at 1 day interval from 0d through 10d). 

I). the sampledata table looks like (tab-delimited),
subject    time (d)
A    D0
...
A    D10

B    D0
...
B    D10

C    D0
...
C    D10

D    D0
...
D    D10

E    D0
...
E    D10

II). The objective is to compare each time point with 0d (baseline reference) within each subject, separately: for example, D_D10 vs D_D0 (not the default baseline A_D0). I tried to set A, B, C, D & E by turn as baseline subjects and D0 as baseline time and use contrast shown below in example for subject D.

#set relevel for subject 
sampledata$subject <- relevel(sampledata$subject, ref = "D");

#set relevel for time factor (default anyway)
sampledata$time <- relevel(sampledata$time, ref = "D0");

#create DESeq data object & define model design
ddsD <- DESeqDataSetFromMatrix(countData = countsdata, colData = sampledata, design = ~ subject + time);

#LRT test with reduced model
ddsD <- DESeq(ddsD, test="LRT", reduced = ~subject); #or ~time (get the same comparison terms)

#resultsNames
> resultsNames(ddsD)
 [1] "Intercept"          "subject_A_vs_D" "subject_B_vs_D"
 [4] "subject_C_vs_D" "subject_E_vs_D" "time_D1_vs_D0"     
 [7] "time_D2_vs_D0"      "time_D3_vs_D0"      "time_D4_vs_D0"     
[10] "time_D5_vs_D0"      "time_D6_vs_D0"      "time_D7_vs_D0"     
[13] "time_D8_vs_D0"      "time_D9_vs_D0"      "time_D10_vs_D0"

#ignore subject effect terms &  use only (baseline) time point comparison terms as below
results(ddsD, name= "time_D10_vs_D0", test="Wald");

Will this work for my objective? I am not sure which 'D0' is being used here (is it from subject A or D?). If this is wrong, how should I proceed with the test? Thanks.

ADD COMMENTlink modified 5 weeks ago by Michael Love11k • written 6 weeks ago by prppmet0
0
gravatar for Michael Love
5 weeks ago by
Michael Love11k
United States
Michael Love11k wrote:

How do you expect to compare a single observation to a single observation ("compare each time point with 0...within each subject, separately")? There is no replication for that level of analysis. I would recommend that you consult with a local statistician on your experimental design.

ADD COMMENTlink written 5 weeks ago by Michael Love11k
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 2.2.0
Traffic: 125 users visited in the last hour