It is your choice what you use as the reference group/condition to which the "fold-changes" are expressed.

Regarding the analysis: I strongly recommend to analyze the LOG fold-changes.

The "ANOVA" should by a "reapeated measures ANOVA" with the two factors "time" (3 levels) and "treatment" (4 levels). If you have an unbalanced design (e.g. not each time point is measured for each treatment) you can use a mixed model with "individual" as random factor (and "time" and "treatment" as fixed factors).

No you should not do the analysis like that because this is an relative experiment that is relative to your control and some changes may be happens due to the incubation period. so you have to do the analysis by considering the Treatment A Time 0 as a control for day zero only not for day 1 or 2. for day 1 consider treatment A at day 1 as a control and so on. Now collect the data for all days and show in a single graph with different time point.

I have no idea about ANOVA.

@Abishek:

The two-factorial (repeated-measures) ANOVA does just what you say. One factor estimates the mean difference between the treatments one factor estimates the mean difference between the time points, and the interaction estimates the inter-dependencies between different treatments at different time-points.

Again regarding my previous suggestion to analyze the log fold-changes: If you fit the log expression values with the ANOVA model, the estimated coefficients for the main effects are log fold-changes (=mean differences between logs), and the interaction gives the log-ratio of fold-changes (=mean differences of differences between logs).

Thank Jochen Wilhelm and Abhishek Vats for the explanation, from your views what I can understand is that:

1.      For each time point, different treatment can be compared relative to treatment A as control.

2.      For each treatment, the effect of time on the induction (Fold change) of gene expression can be compared relative to 0 time point of each treatment.

Now I have a problem ….

Say,

In treatment A, 1st day after infestation the fold change value is X relative to 0 time point of treatment A.

In treatment B, 1st day after infestation the fold change value is Y relative to 0 time point of treatment B.

So will it be right to compare value X and Y because their values are relative to different controls.

My question is:

(a) I want to see if there is any treatment(A,B,C,D) effect on percentage fold induction at day1 compared to 0 time.

(b) Which treatment effect results in highest percentage fold induction in day 1 compared to 0 time.

My view point:

To answer the same we need to compare all the treatment and time points together relative to one control. For example- 0 time point of treatment A as control and find out fold change for all the other treatment and time point. Then the value we get for day 1 of different treatment can be compared to answer the above questions.

Please Share your views.

 Dear Afruza Zaman

Your first point is right that is

your second point is wrong because If you want to do the analysis like that then you have to be sure that their is no relative difference between treatment A at day 0, day 1 and day 2 that is not likely.

@Abhishek vats...For the second point i am saying i would like to see the change in fold change expression over time. How the fold change varies with time.

Good question, me too can I get the answer?

Best thing would be not to use fold changes for the analysis, but just to express the results. By using fold changes in the way you described (whatever definition of fold change you use), you will have data that are artificially with no variance in the reference category, where fold change is obviously 1 (or 0 if expressed in binary log or log), whatever the original data were. So you underestimate the variability of things.

So 1) analyze your expression levels, after log transformation, using the repeated measures ANOVA as suggested. That will gives changes as delta in log; and 2) convert these changes in fold-changes according to your prefered definition of fold-change [that is, convert the delta in binary log or exponentiate it to have ratios].