Hello,
My project focuses on differential gene induction between two stimuli. I want to see if genes are significantly induced across my entire time course and unfortunately I have only been able to detect a small amount when I know there are more.
I want to make sure that my code is executing the task properly. stimulus.final = stimulus- c1 or c2 stimulus.timepoint = timepoint in hours(0,1,2....) Code: DESEQ object:
dds_mat_final<- DESeqDataSetFromMatrix(data_mat_raw_final_matrix,colData=ColData,design= ~ 1 + stimulus.final+ stimulus.final.complete.timepoint)
res_shrinkbl_local <-lfcShrink(res_local, coef="stimulus.final_c1_vs_c2",lfcThreshold=1)
plotMA(res_shrinkbl_local, ylim=c(-3,3), main="Significant Genes Between c1 vs c2", xlab="Mean of Normalized counts between c1 vs c2", ylab="Log2 Fold Change Difference between c1 and c2")
Does using the coefficient "stimulus.final_c2_vs_c1" compare all of the stimuli across timepoints?
Raisa
3 answers
I don't think you need the "1" in the design; I've never seen that in any DESeq tutorial.
What your design will do is compare the betas to the lambdas with the understanding that lots of the in-group variability is not random, but is controlled by the time point.
The other thing you can do with time points is see if there is a linear relationship, and you can use interactions to see if the linear relationship is different in the lambdas vs the betas.
I have had problems with using an interaction term between stimulus and timepoint that was not able to run as well. How can I fix that? I will post the code.
Is it possible that I can compare it using just the stimuli within my colData matrix?
Log in to answer this question.
Can you show your
colDatato understand how the groups are set up?