MicroRNA thresholding problem

[AntonPetrov]

There may be a systematic problem with the microRNA thresholds. For example, this is a precursor that matches its microRNA family according to RNAcentral (probably using an old Rfam model): https://rnacentral.org/rna/URS0000759B7E/9606

But if you run RNAcentral sequence search now it does not match any Rfam families. The current threshold is 99 bits but the sequence scores at 63.7 so it's not reported although both HGNC and miRBase agree that it's that microRNA. The cutoff seems to be set incorrectly.

The interactive report explains why the threshold was set like that:
https://preview.rfam.org/searches/mirbase/MIPF0000894__mir-3180_relabelled.html

The question is why that URS was not included in Rfam seed if miRBase itself thinks it's mir-3180? If it was in the seed then the curator would have seen it in the report and would have set the threshold accordingly.

One way to check for these issues would be to take all bona-fide RNAcentral microRNA precursors and run Rfam microRNA families on them and see which ones do not get hits, as those have bad thresholds.