Analysis and results microarray data for research

Question

[Bruno]

Hello everyone,

I am analyzing microarray and RNA-seq data in patients with diabetic foot ulcers for my research. I am an undergraduate student, and my group doesn't have much experience with these types of analyses. I would like to receive some tips because my results seem to be different from what would be expected.

So, initially, I analyzed data from the Gene Expression Omnibus of NCBI (code GSE80178) following the provided tutorial. For differential expression analyses with limma, I compared expression levels of the control group (non-diabetic patients) with the group of interest (diabetics with DFU).

However, the results of the differential expression analysis returned low values, close to 1.9 in log2FC. Because of this, in order to have a reasonably significant number of genes to be analyzed, I used a cutoff (log2FC > 0.6 or < -0.6; adj.P.Val < 0.05), and I received a list of 274 differentially expressed genes as output. Another factor that raised this question is that when I analyzed the same data using GEO2R, I obtained different log2FC values compared to my analysis using the limma package.

My questions are as follows:

Can a cutoff value of 0.6 be used for differential expression analyses in microarray data? Is 274 genes a very small number of differentially modulated genes for microarray analyses? Is there any reason that could explain these low log2FC values?

[GenoMax]

Is 274 genes a very small number of differentially modulated genes

Remember you are formulating hypothesis for further testing with array experiment so in terms of designing downstream experiments it could be argued that the number may be large.

Answer 1

[jv]

Can a cutoff value of 0.6 be used for differential expression analyses in microarray data?

You can use whatever cutoff you want so long as there's good reason for it - there's no magic number. Also note that a log2 fold change of 0.6 equals 1.5-fold increase in expression (2^0.6 = 1.5)

I don't know what processing steps occur with GEO2R but unless it's exactly the same as code the you used for your limma analysis then it is not surprising that the results are different because different parameters and different tools will yield different results