Please Help Me! Replication In Time Course Experiment

Question

[biosystat]

Dear Biostars,

I have a very simple question!! Actually I have just started working with Mev, it's amazing and user friendly. But the problem is I dont know much about statistics, I mean I am a biology man! I have uploaded my data on MeV, it seems that I should do some normalization, etc. I have 7 samples (time line) with 3 replications, what shall I do with replication, should I just consider one of them as expression data or ? Here is the information of my experiment :

Human Genome U133 Plus 2.0 
Array 40000 genes = 40000 row 
7 sample (time) * 3 rep = 24 column and

a part of data : every three of them are for one sample/time (replication for one experiment )

AFFX-BIOB-5_AT    **7.81507    7.773    8.3**    ,.....

Answer 1

[kanwarjag]

Just to confirm you are using Affy array? I generally use MEV for graphics but agree that you can perform simple t test, anova etc. Assuming it is affy- you have to normalize the data -RMA, quantile ........... you have to see the data. and then go for any statistical test. In general you may also like to see how your replicates are behaving (assuming biological replicates). GE should not be that hard there are various good tools out there.

[biosystat]

Dear Kanwarjag,

Thank you for answer and helps, I dont not know much about data, even dont not know they are normalized or not! All things that I have just a sheet of Excel with rows and column (assume downloaded from public database), the data are between (2- 15), should I still normalize them?

data are from Human Genome U133 Plus 2.0 chip and I found out that it is a affy array... so could you please tell me what you mean by "replication are behaving ?!" and what is" GE "means?! sorry for too simple question !

[kanwarjag]

Hold on you are using publically available data or some core is giving you the excel sheet. If core is giving you the data in excel and it is affy chip (U133 plus 2) it is normalized (most likely by RMA) if you read in the first few rows it generally says. So you can either do pairwise (t test) or ANOVA (> 2 grps).even in Excel (though I would not suggest). When I say checking your replicates- means you have to see if any replicate is showing some odd data (failed or less efficient). Simple scatter plots or Mv plots may be good start.

Answer 2

[Charles Warden]

You'll need some comfort working in R (or a bioinformatics collaborator), but I would recommend the timecoruse Bioconductor package:

http://www.bioconductor.org/packages/2.12/bioc/html/timecourse.html

It can work with a variety of designs. That way, you don't have to worry about whether you technically could have used a simple statistical test (like 2-way ANOVA)