Handling Microarray datasets from different experiments for DEG analysis

Question

[a.lathifbt]

Greetings!

I am planning to work with the gene expression data of T2DM from several different experiments that include several groups within them (all of them are from the same platform though i.e, Affymetrix), to get the differentially expressed genes using Bioconductor packages. However, I am unsure of which approach to use for this analysis. I was planning to get the DEGs from individual datasets separately and then find the common DEGs across the different datasets. Is this a correct approach for this type of analysis? Or do I need to merge the datasets into a single Expression Matrix and go ahead from there?

Thanks in advance!

[a.lathifbt]

Which R package should I use for this type of Meta-analysis?

Answer 1

[JJ]

You could do a meta-analysis. So analyse them separately and then compute a meta-p value and/or a meta-log2-FC.

I was planning to get the DEGs from individual datasets separately and then find the common DEGs across the different datasets.

By just using the common genes you could potentially loose a lot of information - in particular if the datasets vary a lot. Also, you need to account for the different genes interrogated - you mentioned that the datasets all are from the same platform - which is good - but are they also analysed with the same array type? If not, then you need to account for the different number of genes on the arrays.

[a.lathifbt]

Thank you. I will check if the array type is the same across all datasets.

My plan is to use normalize the datasets using RMA and fit a linear model using limma. So what I understood from your suggestion is that after I found DEGs in the individual dataset, I need to take statistics of these genes from individual datasets and do a meta-analysis, right? Can you please suggest an R package to do this kind of analysis...

And further downstream of this, I am working on studying the comorbidity of T2DM with few other diseases (for which I will find the DEGs as well). In this case, Is finding common genes between T2DM (from this meta-analysis) and the diseases in question followed by functional enrichment, the right way to do it?

[JJ]

My plan is to use normalize the datasets using RMA and fit a linear model using limma. So what I understood from your suggestion is that after I found DEGs in the individual dataset, I need to take statistics of these genes from individual datasets and do a meta-analysis, right?

Yes. So you would get an individual p-value and log2-FC.

Can you please suggest an R package to do this kind of analysis...

Hm, not sure if there is a package especially for microarray analysis ... you can have a look at the metafor package.