A look at Simple RNA Switch lab Results

  • 1
  • Article
  • Updated 8 years ago
The question is "What went wrong with the round 1 candidates?"

I've created a Google Document that looks at each of the SRS Round 1 synthesis results and tries to understand what went wrong in each case. Comments, observations, and content additions are welcome to be added to this forum thread.

If this is well received, I may do the same with later results.
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
  • happy

Posted 8 years ago

  • 1
Photo of hoglahoo

hoglahoo

  • 141 Posts
  • 39 Reply Likes
I like your presentation of the images and brief summaries of your thoughts for each design
Photo of Edward Lane

Edward Lane

  • 139 Posts
  • 8 Reply Likes
That struck me as really useful - it still leaves me with a faint hope for some of the blur style designs, and I'll look for more potential mismatches in future rounds.

Once we get the round 2 results back I'd like to see this sort of analysis of those too - this is the sort of feedback I always imagined one might get for the lab results, prior to the next round commencing. To let us hone our designs.
Photo of Eli Fisker

Eli Fisker

  • 2266 Posts
  • 509 Reply Likes
Hi JL!

Read through your analysis of our labresults. Just wanted to say good work, you have a lot of great points.
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
Started a follow-on for round 2 here. A number of near misses, with a lot of the problems seemingly due to unnecessary Cs and Gs (added for color?). Only 3 looked at so far. I probably won't get to many more before round 4 closes - sorry.
Photo of Edward Lane

Edward Lane

  • 139 Posts
  • 8 Reply Likes
Well I timed this question nicely, being vain and seeking to understand the estimate mode for my first synthesised lab result (round 2)

( http://eterna.cmu.edu/eterna_page.php... )

I submitted the sequence
GGAACGGCGAGGAUAUUUUAAAAGAGAGAAGGCGCCAAAAAAAAAGCCUUCUAAAGAAACAACAACAACAAC

the target is obviously
..........................(((((((............)))))))....................

and I propose the estimated shape ought to be different to the suggested estimated shape date

I'm thinking the continuos colour suggests something like this would make more sense
....(...).................(((((((............)))))))....................

So mocking it up in puzzle maker - my proposition would look like this http://eterna.cmu.edu//sites/default/...

And that seems to me a more plausible interpretation of the continuous shape data ( http://eterna.cmu.edu//sites/default/... ) than the suggested expected shape ( http://eterna.cmu.edu//sites/default/... )

The suggested expected shape seems to me to have odd values for nt 26 to 32, and it also seems unlikely to me that 16-19 would pair where suggested given they are UUUU and could pair anywhere from 37 to 45 with AAAA rather than AA-GA if the rest of the mismatching is accurate, but that none of 37-45 shows and lack of availabilty - if the suggested estimate is mostly accurate then the shape data should at least be slightly blue for those nucleotides I would think?
Given that the shape data for those nts is not giving me a value I understand, can someone give me a full (over complex 3D pseudoknot and non canonical base pair) explanation of why the esitmated result is as suggested rather than I proposed ?

I don't mind at all if I'm wrong, I'd mostly like to be right, but I'm just as happy if you can teach me what I've not understood.

Cheers
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
@Edward: I don't even pretend to understand why the estimation mode prefers what it describes over something else. I simply try to understand what could be contributing to it folding as estimated. In your case, the GGC at 6-8 is predicted as folding a strong (-6.7 kcal) 3-pair stack with the GCC at 46-48. (If the system is only looking for pairings within 30 bases, it would miss this possibility.) That would likely be the beginning. After that the 1-4/49-52 mismatch is another relatively strong -5.4 kcal 4-pair stack. Having formed those two, the remaining mismatches are likely opportunistic in that they cause minor drops in the energy of the remaining loops.

At least that's my theory of the moment.

And yes, you are right, a 5-9 C-G mismatch is also possible and likely happens some of the time. In fact with a test-tube full of billions of RNA sequences, they don't *all* fold the same way all the time. Some fold one way for a while, while others fold another way, then something jostles some of them and some refold yet another way. That's why some of the nucleotides are shown as "grey" in the SHAPE results: some of the time, some of them are bonded.

The object therefore is to design a sequence where (1) the most probable structure for the sequence to fold into is the target sequence, and (2) there is a sufficient energy difference between that sequence and other "misfolds" for the misfolds to be of low probability. But being of low probablility does not mean they don't happen at all, it just means that at any given time, fewer of a testube full of RNAs are likely to be folded in that way.

A useful online tool to get a better understanding of this is the "barriers" web server: http://rna.tbi.univie.ac.at/cgi-bin/barriers.cgi
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
A follow-on for round 2 is found here. A number of near misses, with a lot of the problems seemingly due to unnecessary Cs and Gs (added for color?), or similar miss-folds among the Blur variants. Cs in 1-26 were still causing problems, as did strong bonded hairpins that seemed to be forming even without the target molecule.

A textual analysis of all round 2 submissions. Pictures for 3 so far.