We used a GeCKO library to carry out a genome-wide knock-out screen. We obtained our sequencing and ran it through MAGeCK to obtain the raw counts. We've got 2 biological replicates across 2 timepoints, at T=0 and T=8 doublings. I noticed that there was a lot of variability between the two biological replicates, and when I ran MAGeCK MLE, the resulting hit list was quite small and a lot of stuff came out as not significant, likely due to the difference between the two replicates.
I am trying to do a manual analysis of the data to try filter out the most discordant genes and test a smaller set of genes for any significant differences. After filtering out genes where a majority of the guides misbehave, I've settled on a smaller subset of the data, but my question is what is the best way to now test the statistical difference? I have currently settled on a paired Wilcoxon test, but there is there a reason I am not aware of to avoid this test in the context of CRISPR counts?
Grateful for any input