My question in short: Does ConsensusPathDB's enrichment analysis tool accept/work well with median normalized gene counts? The documentation doesn't seem to specify normalization requirements.
Details: I'm inputting median normalized gene counts from EBSeq into ConsensusPathDB's enrichment analysis tool. My input file contains entrez ID's, phenotype 1 normalized gene counts, and phenotype 2 normalized gene counts as three separate columns. These genes are my differentially expressed genes only, not my entire gene list. I have to bump the p-value threshold up to 0.05 to get results for GO terms and pathways. Even then, I only get one pathway and five GO terms. (I selected all GO term levels.) While some of the p-values aren't great (range from 0.00618 to 0.0469), they're tolerable. On the other hand, the q-values range from 0.28 all the way up to 0.945, which concerns me.
I ran my differentially expressed gene list against a background of all my genes with CPDB's over-representation tool and am happy with the results; I got many GO terms and pathways with a p-value threshold of 0.01 and all q-values were all < 0.08. Plus, the results make sense in the context of my research. I just want to make sure that I'm not making a newbie mistake by running incorrect data through CPDB's enrichment analysis tool. Are there other tools that will accept gene counts for enrichment analysis, so I can verify my CPDB enrichment results? (I'm only familiar with other tools like DAVID that accept a ranked gene list without gene counts for each phenotype, and I'm looking for something that will accept gene counts like CPDB.)
I have the same exact question as you. Overrepresentation analysis gave a lot of significant terms. Have you gotten an answer for this yet from other sources?
Try doing a log transformation.