Hello!

I have a set of gene expression data that compares effects of a treatment to control over time, in cells that are undergoing differentiation. I have five experimental groups: time zero (just before treatment was applied) and 12 and 48 hrs, +/- treatment. I want to determine how treatment influences the expected changes over time (those related to differentiation). I can do the 2-way ANOVA with interaction, using the 12 and 48 hr groups, but this misses the effects from 0 to 12 hrs. Is it valid to use the same set of time zero data for both treatments, so that I have a balanced design? Any other suggestions about how to analyze these data?

thanks!

Code should be placed in three backticks as shown below

# include your problematic code here with any corresponding output 
# please also include the results of running the following in an R session 

sessionInfo( )

The experiment you describe is very standard and is commonly analyzed using the methods described in the limma and edgeR User Guides. The basic method is to define the 5 experimental groups and to form targeted contrasts between the groups to detect the different time effects with and without treatment.

I'm a professional statistician, and I was brought up on anova methods, but I strongly advise against 2-way anova for gene expression experiments because the results are difficult to interpret. The ANOVA factorial approach is just too generic and is not focused on the questions that are of biological interest in gene expression experiments.

Is it valid to use the same set of time zero data for both treatments, so that I have a balanced design?

If you mean duplicating the zero time data then, no, that would not be valid, nor is it at all necessary. Having just one group at time 0 causes no problems for the analysis.