miRNAs differential expression with no replicates

Question

[manekineko]

I have read some similar topics but did not find the enswer what is the maximumyou can doo in such cases, and the best way to normalise and make diff.expression.

The things I tried so far:

• doing Deseq, but all p-values as espected are terrible :)
• some people recommend me just to make simple normalization as RPM normalizing by total reads in the library (mir count/total lib reads * 1 milion) and look the folds, but then I read that such normalization is a bad idea.

(I also saw that RPM can be normalized not to total number of reads, but by total aligned to the genome reads or total aligned to miRbase or Rfam reads)

So I'm a bit confused what to do here?

[Sean Davis]

While your p-values are "terrible", you could use them as a ranking of your genes. Any statistical significance test without replication will be pretty challenging to use, anyway.

[manekineko]

What do you mean touse them as a ranking of your genes? Can you explain how to do that? Can i just look the log2foldings to guess how to proseed with wetwork validation?