Models identical in cDNA but not in CDS considered redundant by Mikado prepare

[lucventurini]

Mikado prepare does consider only the exonic structure when discarding models – not the CDS. So if two models were identical in terms of exons but had a different CDS (say, because only one model had its ORF called) Mikado would still consider the pair “redundant”. This is not a problem for how we used Mikado so far (ie on transcript assemblies and/or PacBio reads) but in rare cases might have led to the loss of models in some loci.

To make the type of situation clearer: presume to have four models, A, A’, A’’ and A’’’, all on the same strand and with two exons: (1001, 1500) and (1700, 2000). Model A does not have any CDS. Model A’ has an ORF with coordinates (1201, 1470). Model A’’ has an ORF with coordinates (1401, 1900). Model A’’’ is identical to model A’’. At the moment, as they have the same exonic coordinates, Mikado would consider these four models identical and keep only one of these (chosen randomly).

The expected and desired behaviour is that only one between A’’ and A’’’ will be discarded (as they have also an identical CDS).

[lucventurini]

Properly tested. Closing.