In RNAseq data, I want to test statistically whether a set of 17 genes are differentially expressed in a condition. I have two conditions: control treated/drug treated. My quick & dirty way of looking at it was: I computed up/downregulation vs control by simply dividing the counts in drug treated by the counts in control treated. I thus have 17 fold-change for my 17 genes (eg.
1.2, etc.). I averaged these fold-changes (average is
0.85, so these genes seem to be downregulated as a set in average) To have an idea whether this result had anything surprising, I computed the same average for 10,000 random sets of 17 genes. 90% of the random sets are over
0.85, 10% are below.
What's the formal way of doing this?