Hi there guys. I have gene expression data collected from nCounter with a panel of around 700 genes (Pancancer immune profiling panel). I performed differential expression analysis and now I want to see which immune-related pathways are over or under expressed in my samples. In my nCounter data I do not possess non-tumoral samples, therefore I downloaded a dataset that consists of several datasets of tumor and non-tumor samples joined together by batch correction using RUV-normalization. The expression data is in Log2 RUV-normalized format. From this dataset I only selected the genes that are present in the Nanostring panel I used for my samples.
So, in this data I tried to use both the fgsea and clusterProfiler packages in R to perform this analysis, but neither provided me with understandable results. I either get non-significant results, or only one significant results or even significant p-values but non significant ajusted p-values...
I am wondering if the fact that I am using a small panel that is already "enriched" for immune genes can have an impact on my results. Also, I do not know if the fact that my data is in this format is also relevant for these results.
Should I define the background genes of the fgsea() function to the genes present in my panel? If so how can I do that?
Hope you can help me, Thank you in advance