I am a complete beginner in bioinformatics and am starting by performing Differential Expression gene analysis. I extracted one of the public studies from the NCBI GEO repository. The study has microarray data for 11 patients, the patients have been administered a drug and depending on their response to drugs they are classified as good responder (G), moderate responder (M), and poor responder (P). Now, I want to see if certain genes are differentially expressed between these 3 groups. I used the GEO2R tool which uses limma for the analysis, I extracted the R script that GEO2R provides to see what it is doing.
I think I fairly understand the code, but I have problems interpreting the results. The GEO2R spits out p-values, adjusted p-values, B-value, etc. I want to see for which contrast/ comparison these results are for, so when displaying the topTable, I include the argument coef and set the value as 1, 2, or 3 depending on the contrasts I tested (G-M, G-P, M-P). I find the top table values for each of these contrast do not match the value GEO2R spitted out at the start where coef is not specified. I am wondering for which contrast is the GEO2R giving results for and why does this not match any of the 3 contrasts?
What I also did was I instead of including 3 factors, I selected only 2 factors (eg G vs M) in GEO2R and performed the analysis on these 2. I then loaded the R code and checked the method and obtained the top Table, what I find is the top Table results for G vs M in this does not correspond to the G vs M results from the previous analysis. Interestingly, by looking at the adjusted p-values and assuming significance level is 0.005, this results gives 0 significantly differentially expressed genes, while from the previous analysis there are actually 9 significantly differentially expressed genes.
This discrepancy is really really confusing. Am I interpreting the results correctly? and why does the same data give different results?