Hi, Joke. Thank you for your answer. I have 3 different conditions for each group of animals and time course. Could you recommend me also good facility with reasonable prices?  

Do you have independent controls for each of your 3 conditions and time points?  Do you have good replication (min. 3 biological replicates, but 4 to 5+ is much better)?

Either technology could be used for such a study.  It might come down to cost and availability of services for you.  If wanting to do NGS, you will need to get a cost estimate for minimum coverage, which for differential expression should be a bare minimum of 10M reads per sample, but 25M or more would be extremely desirable as you will miss fewer rare transcripts that way and have much better discrimination of relative changes with 25-50M reads per sample.

When comparing different conditions, one often used approach is to analyze each condition independently, and then compare the results between them (for example, non-parametric analyses based on independent fold change of each condition).  However, this is only possible if you have independent controls for each condition.

Directly comparing two treatment conditions, regardless of array or NGS data, is problematic as it is difficult to interpret what it really tells you about the effects of each condition.  That is, if you see a gene that is up-regulated in condition 1 relative to condition 2, what does that mean if you have no idea about condition 1 independently?  On the other hand, if you see an independent statistically significant up-regulated fold change of X with condition 1, and an independent statistically significant up-regulated fold change of 2X in condition 2 then you at least know something about the relative differences in response by condition in that gene.  And you can then separately test if that observed difference in the magnitude of response is also significant (you could use RankProduct of fold change, or z-scores or some other test).

In the extreme instance where no biological replicates were available, my preference would be specifically for Affymetrix cartridge arrays (with both mismatch and perfect match probes).  In the absence of replicates, there are a couple of robust statistical tools to take advantage of simple pair-wise array comparisons that leverage the specific design of the PM&MM probe sets (e.g. SScore).

Thank you , Michael for very detailed explanation. Actually, I have 6 rats per time point for 24 hours and also 3 different conditions for each group of animals. So, It's tricky for me to decide.