Comparison: NUPACK, InfoRNA, RNAfold, RNAshapes, CentroidFold & CONTRAfold

Which of these packages is the most up to date? I’ve noticed that NUPACK gives results that are not so far off from EteRNA, but the other’s give very different results for the 2nd state shape for a given sequence. So which one is the most accurate and up to date?

BTW these are all mentioned in the EteRNA wiki.

Hi Mike!

This is not a direct answer to your question on the most up to date and accurate RNA structure prediction software. However you might find this post interesting. Some Eterna players ran tests on different software packages.

How useful is structure prediction software in lab

You mentions Nupack and from what I have seen in lab, Nupack is making better designs than Vienna (RNAfold). It has actually made several winners.

I don’t use outside tools. But I sometimes do run a test when the name of a RNA structure prediction tool pops up in chat. I then throw the tool the sequences of some winners from lab to get an idea of its strength on predicting the structure. I also regularly throw it the sequence of some bad designs to check that the tool is not too optimistic. doesn’t have them all, but quite a few.

So who is using the latest energy model? I saw Brourd say in chat that EteRNA is not using the latest energy model. I want something with the latest energy model.

NUPACK isn’t mention anywhere in there. I like NUPACK because it gives more information.

I’ve noticed that NUPACK is the most aggreable, *sometimes* matching what EteRNA’s dot plots say, but RNAfold and CentroidFold don’t ever seem to agree with what EteRNA says. (The other 4 listed in the EteRNA wiki don’t work.)

Have you seen the list of competing algorithms on that website I linked ? Specially, its length ? What does it tell us ? To me, it’s a very clear sign that no single algorithm, no software, no method, not even a global approach is overwhelmingly better than the others.

Nearest neighbor, partition functions and centroids, maximum expected accuracy, local optimas, mixes of these, etc, everyone will tell you that they are doing better than the others. Truth is, they (and we) are still poking in the dark…

Yes, and I noticed the great length of the entries and even the large number of categories on Wikipedia’s “List of RNA structure prediction software”. I’m trying to go through the list and it is a laborious process. But so far I’ve only found 2 that I like, NUPACK and RNAfold.

Eyeballed the data from Eli’s link above “How useful is…” & it looked like there may be some grossly linear relation in most of the plots for free energy vs. synth score in all samples charted. By this I mean I can hand draw a line that evenly divides the individual sets & they fall into two main categories, those with a horizontal divide, so no apparent relationship, and those with a -10kcal:+10%score improvement (with score between 50% & 100%). Not particularly helpful, but somewhat interesting to me.

Need an AI software scanner to drudge through all the RNA predictors.