**20**wrote:

Dear community,

i would like to address my important questions regarding a complex experimental design in R, concerning an agilent microarray dataset with multiple time points. Briefly, after normalizing, a view of my phenotype information is the following:

> y2$targets Sample.and.Data.Relationship.Format time Batch GSM526606 irradiated 0.5 h 1 GSM526607 irradiated 0.5 h 2 GSM526608 irradiated 0.5 h 3 GSM526609 irradiated 0.5 h 4 GSM526610 bystander 0.5 h 1 GSM526611 bystander 0.5 h 2 GSM526612 bystander 0.5 h 3 GSM526613 bystander 0.5 h 4 GSM526614 control 0.5 h 1 GSM526615 control 0.5 h 2 GSM526616 control 0.5 h 3 GSM526617 control 0.5 h 4 GSM526618 irradiated 1.0 h 1 GSM526619 irradiated 1.0 h 2 GSM526620 irradiated 1.0 h 3 GSM526621 irradiated 1.0 h 4 GSM526622 bystander 1.0 h 1 GSM526623 bystander 1.0 h 2 GSM526624 bystander 1.0 h 3 GSM526625 bystander 1.0 h 4 GSM526626 control 1.0 h 1 GSM526627 control 1.0 h 2 GSM526628 control 1.0 h 3 GSM526629 control 1.0 h 4 GSM526630 irradiated 2.0 h 1 GSM526631 irradiated 2.0 h 2 GSM526632 irradiated 2.0 h 3 GSM526633 irradiated 2.0 h 4 GSM526634 bystander 2.0 h 1 GSM526635 bystander 2.0 h 2 GSM526636 bystander 2.0 h 3 GSM526637 bystander 2.0 h 4 GSM526638 control 2.0 h 1 GSM526639 control 2.0 h 2 GSM526640 control 2.0 h 3 GSM526641 control 2.0 h 4 GSM526642 irradiated 4.0 h 1 GSM526643 irradiated 4.0 h 2 GSM526644 irradiated 4.0 h 3 GSM526645 irradiated 4.0 h 4 GSM526646 bystander 4.0 h 1 GSM526647 bystander 4.0 h 2 GSM526648 bystander 4.0 h 3 GSM526649 bystander 4.0 h 4 GSM526650 control 4.0 h 1 GSM526651 control 4.0 h 2 GSM526652 control 4.0 h 3 GSM526653 control 4.0 h 4 GSM526654 irradiated 6.0 h 1 GSM526655 irradiated 6.0 h 2 GSM526656 irradiated 6.0 h 3 GSM526657 irradiated 6.0 h 4 GSM526658 bystander 6.0 h 1 GSM526659 bystander 6.0 h 2 GSM526660 bystander 6.0 h 3 GSM526661 bystander 6.0 h 4 GSM526662 control 6.0 h 1 GSM526663 control 6.0 h 2 GSM526664 control 6.0 h 3 GSM526665 control 6.0 h 4 GSM526666 irradiated 24.0 h 1 GSM526667 irradiated 24.0 h 2 GSM526668 irradiated 24.0 h 3 GSM526669 irradiated 24.0 h 4 GSM526670 bystander 24.0 h 1 GSM526671 bystander 24.0 h 2 GSM526672 bystander 24.0 h 3 GSM526673 bystander 24.0 h 4 GSM526674 control 24.0 h 1 GSM526675 control 24.0 h 2 GSM526676 control 24.0 h 3 GSM526677 control 24.0 h 4 |

As you can see from the above phenotype data frame, the experimental design is rather complex: i have **6 different time points** of **3 different conditions(control, bystander & irradiated cells)** and also **4 different biological replicates of each condition-time point**, which are called **"batches"** for simplicity. Thus, at a first glance I'm mainly interest to test for DE genes between **bystander vs control **and **irradiated vs control **at a specific timepoint, which is the **4h time point**. So, because I'm a newbie to R and complex methodologies, how should i formulate my design matrix with limma, for instance for this specific example ? In order also to adjust for the "possible batch effect", that is the 4 different biological replicates ?

Here is a link to an MDS plot of my dataset after normalization (where the color represents the 3 different groups of comparison) :

https://www.dropbox.com/s/qv3n5lrnlh7focg/plotMDS_after_normalization.png?dl=0

I found in the limma users guide the tutorial 9.6.2 in page 49, but the additional complexity in my design is also the different biological replicates !!

Thank you in advance !!!

Konstantinos

**24k**• written 3.3 years ago by Konstantinos Yeles •

**20**