Short isomiRs and iso_snv_central_supp

[ThomasDesvignes]

An interesting case was raised up:
Let's imagine a miRNA that is 19nt long and has 3 mismatches at its 3'end. How do we classify it?
=> Should it be iso_add3p:3?
or should it be iso_add3p:2,iso_snv_central_supp?

Each categorization could be backed. But some filtering criteria could solve the problem beforehand:

• what sizes of isomiRs should be allowed? A lot of us tend to allow 17-18nt to 25-26nt as a size range. But I remember Bastian F. pointing out that no isomiRs, if not-degraded, should be below 20nt. Bastian, do you have a citation for this statement?
• How many non-templated 3'-end additions seem biologically acceptable? I personally put a cap at 3nt. What do other people do?

If we say that a miRNA shouldn't be shorter than 20nt and shouldn't have more than 3 non-templated 3'-end nt additions then we solve the original question. But is that biologically sound?