I need to design a few synthetic RNA molecules with a known secondary structure (ideally a single stem loop). I want to use them to test an RNA structure determination method, so it is important that I am confident about how the sequence is going to fold (so that I can use the known structure to verify my results are correct).
My first attempt was a script which generates a random RNA sequence (of my desired length and GC content), predicts its optimal secondary structure using RNAfold, and checks if that secondary structure matches the general pattern I want (single stem loop with no bulges). When it finds a matching sequence, it uses RNAsubopt to get all the suboptimal predicted structures for that sequence. It then outputs only sequences where the dG of the second-most optimal structure is far enough away from the first-most optimal structure (say 3 kcal/mole) to be 'confident' that only the optimal structure will form.
I realize this method is probably horribly inefficient, but I do not know enough about RNA biology nor structure prediction methods to figure out how to improve it. It does work when I relax the constraints (e.g., requiring dG difference of only 2 kcal/mol, or allowing stem-loop structure with bulges), but no sequences are found with my requirements.
Is my hope (to design a single-stable RNA molecule) unreasonable? Or is my method just useless? I tried looking in papers to see how others do this, but they don't usually specify the method. I would appreciate any tips or a nudge in the right direction.