Kissing loop riboswitches

  • 1
  • Article
  • Updated 2 years ago
This is a thread for discussing the analysis of the new riboswitches that uses different aptamers and kissing loops:

Single-input switches, revisited (with RNA kissing loop probes): http://www.eternagame.org/web/lab/7630523/


Initial data are available:

Google Spreadsheet: https://docs.google.com/spreadsheets/d/1Vq5vCtGjSH-qSj2Fbw0sZyiN_vt0e1odyhqkOB45AVc/edit?usp=sharing

Excel file: https://drive.google.com/file/d/0B_N0OA9NROPGbC1ucjJTR1dNUk0/view?usp=sharing
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes

Posted 2 years ago

  • 1
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
For those who find fusion tables useful for browsing the data, I've created one at https://fusiontables.google.com/data?docid=1ebFE5HKgFcr3RxtOzKv4Fy1Enxv8PSr4D7NzOIV7#rows:id=1 .  It includes a column for the switch graphs, but as of this writing, those are not yet up on Amazon.
Photo of whbob

whbob

  • 193 Posts
  • 58 Reply Likes
At first glance, Tryplophan B and FMN B (Same State) scored higher ahead of (Exclusion).  How would KDOFF and KDON be calculated for these puzzles? I mean what KD columns could represent KDOFF and KDON?  
Photo of whbob

whbob

  • 193 Posts
  • 58 Reply Likes
Pi's design 7554094 score 92.36 has a KDOFF of 1610.64.  There is no other column with that number. no ligand K4 is higher.  AM I correct the higher the number the darker the measurement? 
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
I missed your previous question about how KDOFF is calculated, but now that I see it, something does look odd.  I have always understood that KDOFF represented the KD value for the state that was expected to more weakly bind the reporter. For a Same State puzzle, that would be the state with no ligand. So I expected KDOFF, by definition, to be KD_noligand.  But as you have pointed out, that is not the case for the design you highlighted, and I have confirmed that it is not true for any of the KL SS solutions.

I'll ask Johan if this is intentional.
(Edited)
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
@whbob If you are referring to puzzle 7754094, KDOFF comes from KDnoligand_K4 and KDON from KDTrp_2_4mM_K4. The discrepancy in values comes from normalizing values from averages in dG vs KD space. It could perhaps be fixed but the values are close.
Photo of whbob

whbob

  • 193 Posts
  • 58 Reply Likes
thank you Johan, so with no ligand even with FMN at 200nM the signal is dark.  When tryptophan A is added to 4 nM the signal brightens?  
np with the discrepancy, just  wanted to confirm the columns used.
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
I updated the excelsheet with well-normalized values. 
Here are the google sheets
R107 results, kissing loops
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
I made a fusion table of the updated spreadsheet and merged it with Omei's fusion table to get the switch graphs in there as well.

R107 results kissing loops - Normalized + switch graphs
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
The MS2/FMN riboswitch blueprint still hold up also for other hairpins and aptamers

In the Same state labs - the aptamer likes to be across the kissing loop. Just as the FMN aptamer liked to be across the MS2 hairpin.

Example from best scoring same state design this round versus overall Same State blueprint
  • New thing for this round is that the aptamers in general seem to want to be closer to the kissing loop hairpin compared to FMN versus MS2. So the loop between aptamer and hairpin isn’t as big as for FMN and MS2.


Design by PI, score 92%

https://drive.google.com/open?id=0BxJGPGrJ3fkKU0NPelhTVTNEQnc


The blueprint for the Same state - Tryptophan A (Top half) example is just rotated when compared to the overall best blueprint for our earlier Same State FMN/MS2 riboswitches. (Bottom left)

The blueprint and the static stems are just not placed the same way. But the basic stuff as aptamer and hairpin position is alike. The aptamer and the hairpin holds the exact same position opposite each other and preparing for a hug.

This do not go for all top scoring designs, but this do goes through a lot of them.



Symmetric switch elements


Devs asked a question alongside the lab round conclusion: “The arginine and theophylline were not as good and we hope to work with the Eterna players to understand if it because of design challenges, or if the aptamers are less robust than suggested by previous reports”.


I think this is because of the arginine and the theophylline aptamers are being less symmetric. I already suggested that I thought symmetric aptamers would do better, before we had run the labs. Under the section Aptamer I mention that I think that symmetry for switches also apply to switch elements like aptamers.


Tryptophan is symmetric when not split, in the ON state. But Tryptophan when split, goes more asymmetric.


FMN is supersymmetric not just in structure but also because it hold palindromic sequences.


Palindromic sequences are really just symmetric sequences also. So switches do not just seem to like symmetry on a structure level but on a sequence level as well.


The MS2 hairpin also harbour near symmetry and the Kissing loop definitely has the symmetry.


And when I mention symmetric aptamers, I’m primarely thinking about their bound state. Although FMN can be kind of symmetric when in its unbound state too. This actually happens a lot in same state switches.


Basically when viewing how the FMN binds and unbinds, it is a miniature image of the same movement that the whole design does to switch state. In a crossed manner. It is actually the exact same that is going on the switch element scale as on the design scale.


Exclusion labs


As for Exclusion labs the aptamer still likes to be real close to the kissing loop hairpin, like FMN to MS2. 


Top scoring FMN Exclusion lab design with kissing loop hairpin.


Design by AndrewKae, score 83% (right)


https://drive.google.com/open?id=0BxJGPGrJ3fkKUGRGLTVIaXNrM2M

Earlier exclusion blueprint (left) taken from here. While the angle of the static stem inside the switch design are different, the overall positioning of the kissing hairpin loop to the aptamer is akin to that for FMN/MS2. (Left)

The kissing loop hairpin can in principle be on either side of the aptamer. Which will be best, will depend both on the aptamer but also the switch element hairpin. For FMN’s and MS2’s, the opposite placement was best as I called out in this blueprint overview.
(Edited)
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
On a curious note


For a good deal of the labs I did an experiment with adding double aptamers. The overall best same state designs for MS2/FMN switches, were designs with double FMN aptamers.

I added double aptamers in two different ways.

1) I put them right after each other, just as the original double FMN designs.
2) I reversed the aptamer I added in as extra in relation to the mandatory one.

I did the later for the following reason. Some of the aptamers were asymmetric. I believe symmetry to be central to switches. If I put two asymmetric aptamers right after each other, even more asymmetry would arise which I thought to be not good.

However if I reversed the sequence of one of the aptamers, I would re-create a certain amount of structural symmetry.

Score 86%, double aptamer reversed


https://drive.google.com/open?id=0BxJGPGrJ3fkKMlBMck5fYThxUjA

When I look at what same state designs with double aptamers that comes out on top. It is a tryptophan mod by Eternacac of one of my designs, where I created symmetry for the design with asymmetric aptamers by reversing one of them. It has the third highest Same state score.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
When I mentioned symmetry in the above image, to be more precise I mean there is rotational symmetry. (Thx, Khan Academy!)
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Omei mentioned something that I think is important for us to know to better analyse this rounds data. Here comes the part of our conversation that I deem relevant.


eli: I have noticed something funny. Not sure what it means. The designs with the absolute low KDON for the kl loop are the ones that are full moving. [Or close to full moving] I haven't checked all labs for this yet, but this was what stood out.

omei: Here's something to keep in mind as you look.  There's been discussion among the devs as to how 3D spatial conflicts may influence the CRISPR labs.

I'm wondering whether this might also be a factor with the KL puzzles, because the report in this case is a substantial hairpin loop.

The sequence is GACGCCUGACAUCACCAGGCGUC

eli: So this is the sequence that is pairing with the kl loop?

omei: Yes.

Presumably, if it binds, it has an 8-base helix protruding back toward our design.

eli: Oh, that wouldn't be good. So this would give an advantage to long kl hairpin loops... :)

omei: It certainly has the possibility of interfering with the otherwise normal foldings.
Yes, I would think so.  But no guarantees.

eli: Hehe. I did notice another oddity. A couple of Jieux well scoring mods, had a ton of GU's in a very long switching stem. In tryp B exclusion

omei. That meshes with something else I have noticed -- the range of fmax values is larger than usual.

eli: I still have a very flimsy understanding of fmax values.

omei: I think what is happening is that the reporter is less selective about where it binds.

Ideally, fmax will always be in the range of 0-1 (assuming only one instance of its binding site), and it represents the proportion of designs that will bind to the reporter when the reporter concentration is extremely high.

If there are 2 reporter complements, designs should be able to get an fmax score of 2.

eli: Oh, like double fold change for double FMN aptamers :)

omei: There are always some outliers, but those can usually be explained away as measurement errors.

Similar, yes.  Double FMN aptamers shouldn't be able to increase fmax, but fold change, yes.

Anyway, the fmax values are up to about 3 for these KL designs.

Eli: So is fmax a meassurement of how many kissing loops are able to bind up with a high concentration of that fishing loop sequence?

Oh, I think I get it. You are saying that the kissing loop probe is way too happy binding up where it really shouldn't?

omei: Yes, that's basically right.  Each reporter molecule adds a bit the light intensity, so the more the brighter. Yes, that's my suspicion.
 

So what Omei mentioned about the probing kissing hairpin loop stem potentially protuding back towards our designs, makes me wonder if kissing hairpin loops do not only kiss with their loops, but if they are actually capable of doing coaxial stacking with their stems too?

If so this would be an extra reason for long K4 hairpin loop stems, to be beneficial. Despite the growing penalty for switching stems, the longer they get. As it is easier getting a shorter stem actually switch.
(Edited)
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
For those who are interested in how RNA forms in three dimensions, here's a schematic representation of an in-silico estimation by RNAComposer of how the KL reporter might bind with UUUGCAGGGCAGUGAUGUUGCCCUGCUUU, which was one of Johan's controls (K4-8bp_control).

The flat ribbons represent the RNA backbone, while the "sticks" represent the nucleotides.  A two-color stick indicates the two bases are pair bonded.

Here's the same prediction, at atomic resolution.

The thin blue lines between paired bases indicate the hydrogen bonds that occur when bases pair.

The visualization app is Chimera, which I highly recommend to anyone interested in what is actually going on in three dimensions.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
For those curious about Chimera, Nando and Omei made some fine intros. You can find them under the Tools section on my profile.
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
Here are a more detailed analysis document with lots of plots. 
https://drive.google.com/file/d/0B_N0OA9NROPGcjRxUXFpc0ZFQjQ/view
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes

EXCLUSION LABS


The easiest way of getting a fine score for a riboswitch lab - with an aptamer and a switch hairpin - is to make a switch bubble. That goes for both Exclusion and Same state labs.


With switch bubble I mean having the switching hairpin and the aptamer sequence inside a loop - hence the switch bubble. That way the elements are close for contact when needed.


However some labs can’t handle a switch bubble, due to the positioning of some of their switch elements in relation to each other.  



Kissing hairpin between aptamer sequence (Overall A puzzle type)


  • Exclusion Arg A, Theo A and FMN A keep the KL outside of the aptamer sequence.


Resembles the earlier labs Exclusion 1, Exclusion 3, Exclusion NG 1 and Exclusion NG 3 that had same configuration for its (FMN) aptamer and the switching hairpin (MS2). (Riboswitch on a chip lab series)


As seen for the above mentioned exclusion labs, exclusion designs tended towards having more switching basepairs and bigger switching regions when compared to Same state labs. Exclusion (OFF switch) labs also tends to harbour more fold ups than Same State labs.


What characterize labs with this configuration:

  • Harder to solve

  • Towards open ended solutions

  • Towards fuller moving switch

  • Many switching basepairs


Still there were something extra that made these labs perform worse than they had to. In the above mentioned exclusion labs, a switch bubble did not do well. These labs had a specific characteristic, that I think caused the switch bubble strategy fail.



The aptamer in Exclusion FMN A is turning the wrong way


Generally FMN likes it best if it can have the two set of double G’s at the static end of its aptamer. At the neck of the switch bubble. It likes turning a specific way.


This is exactly what the aptamer can’t do in these labs: Riboswitch on a chip labs: Exclusion 1, Exclusion 3, Exclusion NG 1 and Exclusion NG 3 plus from the current round: Exclusion Arg A, Exclusion Theo A and Exclusion FMN A.


However the Inversed Exclusion NG1 and Inversed Exclusion NG3 lab from Riboswitch on a chip labs, had the FMN aptamer pointed in optimal direction, scored better than their sibling labs that were forced toward doing an open ended switch by their unlucky placed FMN aptamer. Plus these exclusion labs with the inversed and correct turning FMN aptamer, preferred a switch bubble too.


The FMN aptamer can open in both directions, but it has a strong preference for the one side. And when forced to do it the other way around (like in this exclusion FMN A type design below) it prefers an open ended style over a switch bubble.


I won the Exclusion FMN A (with the less optimal turned FMN) lab by doing an open ended design.


Score 80%

https://www.eternagame.org/game/browse/7657282/?filter1=Id&filter1_arg1=7702641&filter1_arg2=7702641


Results sorted by having minimum 50 cluster counts for K4, and a maximum fold change error of 1.25


The winner in the sibling lab Exclusion FMN B type, had the FMN optimally placed for doing a switch bubble.


Score 80%

https://www.eternagame.org/game/browse/7657293/?filter1_arg2=7694708&filter1=Id&filter1_arg1=7694708



How to get better scores for next round of riboswitches :)


Now I wonder whether the Arg, Theo and Tryp aptamer (when split) are showing themselves from their best side, when it comes to forming a switch bubble? Or do they need to be inverted too?


Yup! This goes for the Argenine, Theophylline and FMN puzzles of this round. They need inversion. The Exclusion A type holds its aptamer opposite that of the Exclusion B type, when looking at the option of forming a switch bubble in the puzzle.


So I basically think the Exclusion A type puzzles will have their score benefit from having the aptamer reversed.


The tryptophan aptamer turns the exact same way in relation to the switch bubble, both in the A and the B type puzzle. And it is likely already showing itself from its nicest side.  



Kissing hairpin outside of the aptamer sequence (Overall B puzzle type)


  • Exclusion Arg B, Theo B, FMN B keep the KL hairpin between the aptamer sequence.


Resembles FMN/MS2 Exclusion 2, Exclusion NG 2. And just like these similar FMN/MS2 riboswitches, the best scoring designs prefer a switch bubble.


What generally characterizes these kissing loop labs

  • Switch bubble

  • Fewer switching base pairs compared to the Exclusion A puzzle type

  • Partial switches



Afterthought


I get the sense that the kissing loop labs are less particular about if the switch hairpin is between or outside the aptamer sequence, compared to the FMN/MS2 switches. While the blueprint in the latter labs could be rotated around, they much preferred having the MS2 hairpin between the aptamer sequence. I suspect the latter is due to the MS2 being so strong that some help from both sides were welcome to help split it. Since the Kissing hairpin is weaker, it can easier follow blueprint, no matter which way it’s permutation turn in the switch bubble.


The kissing hairpin loop labs may actually prefer doing the riboswitch blueprint, standing on their head. Having the kissing hairpin loop outside the aptamer sequence. They just want their aptamers turning the right way. :)

(Edited)
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Invert the Same State Aptamers Type A too

The Same State type A and type B puzzles, don't either have the same aptamer orientation. (Except for tryptophan).

I think it is the A type again that needs the inversion.

The Same State A type labs (Arg, Theo and FMN) tend to go toward open ended solving style, something which mostly belongs in an exclusion lab. I think it is the aptamer not placed at the optimal orientation, that is causing it.