Inconsistency between inosine_m6A model and DRACH_m6A model

[wangziyuan66]

Hi,

Currently I am suing [email protected] m6A_inosine and m6A_DRACH models to screen the cell line. I found that there were a lot of non-DRACH m6A sites when I used m6A_inosine model. Can I trust these sites? Or they are just false positives? If they are just FPs, I cannot have the conclusion that the motif of m6A is DRACH.

Best

[ArtRand]

Hello @wangziyuan66,

The 6mA_Inosine model is an all-context model, so all adenine residues will have a base modification call associated with them. All models have a false positive rate, so I would expect that some non-DRACH motifs will still have modification calls. You can see the current RNA modification performance numbers on this Epi2Me blog. If you want to only look at DRACH motifs whilst using Modkit, in pileup for example, add the --motif DRACH 2 option to your command.

[wangziyuan66]

Thanks for your reply, Art! I also find that if I make a seq logo plot, the results are not DRACH if I use all context model, which makes me a little bit confusing. Indeed, I do not see any sequence pattern. Does that mean there are a lot of non-DRACH sites in the cell line (HEK293T).