Does anyone know if the MFE calculations done in the game are strickly algorithmic or do they incorporate any empirical data for special cases? I was wondering because I’ve been doing some outside reading and I’ve been exposed to the whole cis/trans and WC/non-WC [Watson/Crick] universe and UUCG tetraloop stable configurations (among other topics).
I don’t think the EteRNA MFE engine accounts for non-WC pairs [based on other forum posts] and that’s fine but wondered if empirical data was being used for empirically observed patterns/motifs found in nature within the MFE engine.
Thanks.
P.S.: There’s an older post by dimension9 that seems to catalog tetraloops very nicely. I’m still trying to work through that thread. Very impressive work I must say.
EteRNA is currently entirely based on ViennaRNA 1.8.5 (for RNA folding calculations only, of course). ViennaRNA 1.8.5 is fundamentally based on a model called “nearest neighbor”, with a large collection of parameters published by the Matthews Lab, and known as “Turner 1999”, which are described in detail in http://www.researchgate.net/publicati…
To answer your questions: empirical/natural data is used, yes, some noncanonical interactions are taken into account (GA mismatches for instance), and special cases are included (some tetraloops). All of it is still very largely a statistical approximation which ignores a large number of possible motifs and interactions.
Thanks! This also helps explain why I get similar fold predictions when I use NUPACK and select the “Matthews et al. 1999” RNA energy parameter option.
On the topic of the GA-mismatch example, I’ve recently just been thinking of it as a non-WC match where they’re just separated spatially a little more than usual since they’re both purines(?). [That’s just my musing on this, no real data to back it up].
Anyway, thanks again. I’ll be taking a closer look at the PDF you referenced. I just glanced through it and I think I’ll need sit down with a warm beverage to go through it.
GA mismtaches are typically the so-called “sheared G-A” pair, meaning that the Hoogsteen edge of one of the base has hydrogen bonds with the sugar edge of the other. Spatially, it’s not really comparable with pairs connecting by their Watson-Crick edge.
This will give you an idea of how I am exploring your question. It is what got Rhiju’s attention and started the current round of discussion on player software scripting.
This shows how EteRNA does not properly handle a particular known motif, but comes close.
Interesting. I hadn’t encountered yet the unfolded pseudo-knot free (rainbow) graphs of sequences with Leontis/Westhof(?) annotations. It looked like the GC,CG cap on the end (nt #11-#14) wasn’t on the graphs so it threw me off at first. [Either that or I need to get some sleep :-)].
In any event, thanks for the link. It’s an interesting approach.
