Microarray Data Anlysis For Clinical Parameters

Question

[mandar.bobade60]

Hi,

I have normalized microarray data with me. For this normalized matrix, I have information file also which contains information about each sample, i.e. ethnicity, sex and age of each sample or subject. Now, I have been asked to provide result tables based on parameters, mentioned in infomation file.

Here, I am just seeking for help in context of result tables based on parameters. So, what results should I provide based on parameters? Below is the sample information file:

Sample Ethnicity Age Sex Condition

C1A Asian 50 M Malaria

C2A Asian 70 M Dengue

P1A Africa 60 F Malaria

P2B Australia 50 M Dengue

E5C American 38 F Malaria

---

Thanks

[Devon Ryan]

I hope you have more samples than that, having only 5 with so many combinations would make life difficult.

Firstly, you should just ask whomever asked you to provide the results what he/she wants rather than asking a bunch of people who can only infer it. Secondly, yes, presumably the person wants the interaction effect of disease and gender/age/ethnicity (separately, I should add). You can do that in limma. I assume that effects such as Male vs Female aren't important, since you're looking at disease.

[mandar.bobade60]

Yeah, you r right. I have likewise 70 samples, I have just mentioned first 5. So in interaction studies, any hint how I can make groups for analysis.

[Devon Ryan]

Ah, 70 should be enough to look at some interesting effects :) If you're not familiar with limma, have a look through the vignette, particularly on how contrasts work, since you'll be making heavy use of them.

[mandar.bobade60]

Here, what I am thinking is, analysis between multiple combinations, such as Male Vs Female, age above 60 Vs below 60, ethnicity Asian Vs rest or Ameriacal Vs rest. Is this right?

Answer 1

[Sean Davis]

What you are describing is a differential expression analysis. However, since you have multiple "factors" or groupings of samples, you will need a software package that is designed to cope with these multiple factors. The bioconductor limma package comes to mind, so I would suggest reading the fantastic limma User Guide. If this is your first time working with bioconductor or doing analysis, I highly recommend you find a local collaborator or bioinformatics person to help you as you work through your dataset.