Hi, I recently used farfar2 to predict some RNA structures, but it seems that the longest RNA sequence that would succeed is beneath around 500-600 nt. When I submit sequences with 1800, 2000 or 7500 nt, they failed with returning "No output silent files were produced in the FARFAR2 step; please double-check your inputs!", and I couldn't figure out why. So I wonder is there a limitation of RNA length or size when using farfar2 to predict structures(which I've searched and not found yet), or I just made other mistakes somewhere else. Thanks a lot!
Yes, RNAs much longer than that are not going to complete even a single structure in the available walltime. Honestly, though, there's an upstream conceptual issue: it's a very rare RNA that has a single well-defined 3D structure across that many nucleotides. RNAs have domain structure akin to proteins -- even the ribosomal RNAs, for examplem, which are in the 1600-3200 range depending on organism and subunit, can be partitioned into roughly 500 nt domains. So I'd claim that you probably should try to find the underlying domain structure of your RNAs (not easy, necessarily, but maybe evident in the secondary structure!) and then predict individual domains.
That said, even 500 nt is very big and your model accuracy will be limited unless you have templates available. What's your scientific need? What modeling accuracy is useful (or necessary) downstream?