I am carrying out an RNA-seq analysis on mesenchymal stem cells retrieved from disease and control state. However, the exp is without replicates (bad idea I know). The analysis is exploratory, to identify potential factors that can be validated in the lab. So far, I performed transcript quantification using salmon and summarised output to gene level using tximport. So now I have a counts matrix. Just to clarify my only real approach at this point is to rank genes based on a transformed or regularized LogFC?
EdgeR claims to be able to derive p-values for single-replicate experiments. however, you would never be able to publish such results in any reputable journal (only one of those bogus journals that publish anything about anything). Obviously, 3 replicates is ideal; better with 4 or 5.
Things to try:
- look at max and min expressed transcripts by normalised counts
- look at fold-change ratios between the replicates (as you mentioned)
I trust that at least these have been processed in the same batch and in exactly the same way. The last thing that you want is a batch effect with a n=2!