This ends up being a little tricky.
The “STEM” lines define constraint energies that help guide residues towards each other such that they form Watson-Crick type geometries. They also do one more thing — if you have multiple chains (you can define the boundaries with the “CUTPOINT_OPEN” tag), the n chains will be connected across n-1 residue pairs using Watson-Crick geometries drawn from a crystallographic model of the large ribosomal subunit.
You can force a noncanonical pairing by:
OBLIGATE PAIR 10 32 W H A
The obligate forces a rigid body connection between the residues; it will be drawn at random from W/H pairs the ribosome in this case. If the residues 10 and 32 are one different strands, e.g., suppose you have a line like:
then you’re fine. Otherwise, the code will pick a random place to break the chain between 10 and 32, and use an energy term to try to keep the chainbreak as small as possible during folding. If you want to specify this chainbreak location (instead of it being random), you can use a CUTPOINT_CLOSED line.
Finally, the A/P specification is a different classification than the Leontis/Westhof cis/trans classification; it is based on the angle between base normals, and I found this more intuitive. Basically, if you assume the nucleobases are attached to the backbone by anti- chi angles (the “usual” glycosidic angles for RNA), A or P means the backbones are going antiparallel or parallel, respectively. [In the future, we should probably put in cis/trans as an option for specification as well.]
Hope that helps.