How to assess the change of a set of genes in Time x Treatment

#1
Hello,

I have expression level measurement of 40 genes at Baseline and Week4 for the patients had Treatment A and Treatment B. What test should I use to assess if the overall change of these 40 genes is significantly different between A and B from Baseline to Week4?

Thanks!
Grace
 

fed2

Active Member
#2
probably t-test on the change from baseline, or log fold-change. thing is what do you mean by 'overall change'? does that mean any is different?
 
#3
probably t-test on the change from baseline, or log fold-change. thing is what do you mean by 'overall change'? does that mean any is different?
Thank you for your reply! I actually have two models but don't know which is more appropriate.
Model 1:
(1) Calculate the fold change at Week4 from Baseline for each gene and each patient
(2) Calculate the Mean of (1) over the 40 genes for each patient
(3) Compare Treatment A vs. Treatment B groups on (2) via t-test

Model 2:
(1) Calculate the GeometricMean of the 40 genes for each patient at Baseline and Week4, respectively
(2) Use two-way ANOVA (Time X Treatment) to assess if the difference exists between Treatment A and B on the GeometricMean.

Any advice?
Thanks!
 

fed2

Active Member
#4
i think take logs in either case. thats a good question. I would go with '1', but i think they should give roughly similar results. How close id have to get out my anova book. Im assuming procedure 2 is giving 40 p-values, as is model 1?

I think bioinformatics types basically run t-tests as you describe above and then apply post-hoc adjustment, i think 'false discovery rate' control is still in vogue.

sorry im not real big on bioinformatics.
 

fed2

Active Member
#5
on second thought model 2 would have to be repeated measures wouldn't it? so i think (1) is it, final answer.
 

fed2

Active Member
#7
oh i see. well there is nothing which forbids averaging over the genes, but it is just a matter of what is meant by 'overall difference'. sort of depends on science context id think. Im still entering (1) as that was my final answer. isn't it stll repeated measure for (2)? every person contributes baseline and week 4 avg n'est pas?