Hi this is my first post so apologies if the wording is poor.

Disclaimer - (I know any bioinformatics without replicates is not a great way to go, however, I am an honors student and have been asked specifically to work with what I have... (Corona has put a halt on replicates).

I am comparing two initial RNA seq samples and I have been asked by my lab to do some over-expression analysis

I used feature counts to get raw reads then NOISeq in R to simulate biological replicates (pairwise comparison without replicates) and got an average fold change between my condition and control.

I have now been asked to take the genes with the highest fold change and see if they are overrepresented in any pathways (GOseq etc).

This is fine, however, there are certain genes in the control which had a raw featureCounts read count of 1 and where much higher in the condition sample and vice versa thus there exists a large fold change between some of these genes.

*My question is should I assume that the raw expression levels from featureCounts when approximately equal to 0 (i.e 0-2) are truly low expression values and thus differentially expressed between condition and control OR could these low value simply be noise and true values are probably similar between case and control.

Thus should I just not include these genes in my analysis??

below is a rough example of what I'm talking about

guide 1 cntrl

0 42

0 41

0 35

0 24

0 23

1 19

1 16 etc.