mRNA representation

[Hrovatin]

After reading the paper I am unsure if this model can be directly used for mRNA representation without fine-tuning.
Namely, the training data contains no mRNAs (if I understand correctly). However, the model was then FT on mRNA tasks.
1.) May I ask why mRNAs were not used in pre-training - what is the reason for that?
2.) Did you check masking performance on different sequence types including mRNAs? Similar as for structure prediction, although mRNAs were also missing there.