Riboswitches with new aptamers (tryptophan, theophylline, arginine)

  • 1
  • Article
  • Updated 1 year ago

This is a thread for discussing the analysis of the new
riboswitches that uses different aptamers as inputs and MS2 or kissing loops as
reporters. The aptamers in question bind tryptophan, theophylline or arginine, respectively, which presented new challenges compared with earlier puzzles using an aptamer for FMN. Looking forward to the discussion.

 There is also a separate thread for the kissing loop puzzles: https://getsatisfaction.com/eternagame/topics/kissing-loop-riboswitches


The puzzles were posted as two separate Eterna labs:

Single-input switches, revisited: http://www.eternagame.org/web/lab/7559749/

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

Initial data are available (both kissing loops and MS2):

Google Spreadsheet: https://docs.google.com/spreadsheets/d/1KmUbRSYh8pmxa9myHHrLxMfo0nqtpt_EdBzTBVTNIRg/edit?usp=sharing

Excel file:https://drive.google.com/file/d/1jqFmZPEkDM0K_rvkxxoNDG8_RBLSpJBH/view?usp=sharing

Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes

Posted 1 year ago

  • 1
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes
Here is a fusion table with the data from Johan's spreadsheet.

Results R107 - Data: Fusion Table + Switch Graphs

Heads up. If you can't find a design by its designID in the lab where you think it belongs, there is a reason. The data are from two bigger and seperate labs, but some of the sublabs have the exact same names.

There is an extra column in this data collection called Puzzle_Type. Use this one as a filter and that way you can get the kissing loops labs split from the ones with MS2.
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
Thank you Eli for the clarification. The puzzle names were indeed identical between the Kissing Loop and MS2 labs, reflecting the identical puzzle configurations.
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes
The switch graphs were not showing in the first fusion table I shared. My apologies for being new with making them show. Thx to Omei for teaching me.

Results R107 - Data: Fusion Table + Switch Graphs (corrected)
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes

Overall trends

  • The MS2 hairpin works better than the kissing loop hairpin for all but tree labs

(Same state - Argenine B, Same state - Tryptophan A, Exclusion - Argenine A. In Same state - Argenine A, there is a draw between MS2 and the kissing loop hairpin.

Transfer of the FMN/MS2 riboswitch blueprint

The riboswitch blueprint as seen in the FMN/MS2 riboswitches is also present in the labs that holds new aptamers and one of the lab rounds holds a kissing loop hairpin (K4) instead of a MS2.

This can be seen in the following things

  • The puzzles of the B type (that has the aptamer optimally oriented towards the switch area) generally can be solved having the switching elements inside a switch bubble.

  • The best Same state B type designs tends to have the MS2/K4 distanced to the aptamer - as were the case for the FMN/MS2 riboswitches

  • Main part of the best Exclusion B type designs have the MS2/K4 distanced to the aptamer as were the case for the FMN/MS2 riboswitches.

There are more exceptions though, which I find interesting. However these exceptions do turn up in particular in labs where we have a hard time hitting a decent score. Something which were also the case in some FMN/MS2 Exclusion labs. But in those Exclusion FMN/MS2 labs where we had trouble making winners, the FMN aptamer was not optimal orientated and not optimal placed. (more of an A type placement).

In the labs where it seemed hard to get winners, it was seen in exclusion designs that the MS2 was moved further from the aptamer. Which is akin to solves the riboswitch labs, Exclusion 5 and 6. Those were also more likely to be solved in an open ended style due to the aptamer not turning in the optimal direction.


The aptamer is likely turning the wrong way in these labs

As I mentioned in my earlier Kissing loop analysis the following labs probably have their aptamer oriented the wrong way in the following A type puzzles of both Exclusion and Same state kind in both the K4 and MS2 lab type. :

  • Theophylline A

  • Arginine A

  • FMN A

KDON and low scores

  • There is an overall high KDON for the main part of the labs.

  • High KDON means that it is hard for the aptamers to bind the molecule in state 1. It seems hard to get a low enough KDON for the aptamer, to achieve a good score in a lot of the labs.

Average KDON and KDOFF for Kissing Loop riboswitches

Average KDON and KDOFF for MS2 riboswitches

Switch bubble

  • The B puzzle types were more likely to be solvable having the switching parts locked in to a switch bubble where the switch elements are brought close together.  

This conformation was also present in the most successful of the MS2/FMN riboswitches. The switch bubble pattern showed itself in the FMN/MS2 puzzles when

  • The MS2 was put in between the FMN sequence (akin to the B type puzzles)

  • The FMN was orientated in its favored direction

This happened exclusion style labs too if there were enough sequence length to make a switch bubble and the FMN was had its favored orientation towards the switching area.

Open ended

  • The A puzzle labs were more likely to be solved in an open ended style.

This is likely partially due to not all the aptamers having the optimal orientation. I will expect more switch bubble style solves if we saw the round rerun with the A puzzle types having their aptamer orientation reversed. (Except for tryptophan which is turning the exact same way in relation to the switch area for the A and B puzzle types. It is likely turning in the primary orientation already.)

Theophylline goes both ways

One thing I find really interesting is that Theophylline do not seem nearly as picky about what way it is orientated as the FMN aptamer. There are winning designs in 3 of 4 theophylline labs and a near winning score for the last one. Despite the aptamer being inverted for the A puzzle types and the designs being solved in an open ended style.

We have scores of 100 in both the Theophylline in both the A and B puzzle type, despite the orientation of the aptamer in relation to the switching area is not the same. Actually I can’t even say that. Because the Theophylline Same state A lab have a winner in an open ended style solve with an inverted aptamer. So despite the aptamer turns the opposite way to the B type puzzle, the aptamer is actually orientated the same way in relation to the switching area - which this time just is out. Not inwards inside a switch bubble. This is new and very interesting.

Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes
I didn't got it mentioned, but the spreadsheet I shared above I made with a filtering of max 1.25 in fold change errors.

Different cutoff KDON for different puzzles

I am working on a turtorial. Here comes a question and answer section from Omei and I. (8 March 2018)


eli [7:49 PM]
This is one of my candidates
It is by PI and is the best in its lab
It is an exclusion design, but do something rather different
It doesn't have the MS2 next to the aptamer as is normal procedure in Exclusion labs
Instead it has a special structural characteristic.
The shape between the states is almost inverted.
I have seen this before
Also in a switch that did surprisingly well
Can't recall where as of now, but think it was one of JR's. It did a swap over the middle instead. So the portion to the left in one state went to the right in the other state
I suspect we can use this approach

eli [9:03 PM]
Sigh, there is one thing I don't get about the numbers for PI's design. The KDON seems higher than is normal for recieving a full switch subscore.

eli [9:03 PM]

eli [9:03 PM]
I highlighted the more normal value with green

omei [11:02 PM]
The switch subscore is determined only by the fold change, not KD_ON alone.  If you look at KD_OFF for this design, it is also higher than normal for this puzzle.

eli [11:04 PM]
Ok. But I recall that for one of the sub scores that KDON had be below a certain treshold (edited)
I think it was something like 15, but I can't recall for sure

omei [11:04 PM]
That's the baseline score.

eli [11:04 PM]
Ok, then it is that one I am wondering about
I have found another one of Pi's design of the same type that works but have a more reasonable KDON

omei [11:05 PM]
If you sort by KDON, you'll see that there is a cutoff.  But I don't think Johan has used the same value for all puzzles.
It's probably dependent on the affinities of the input and output.

eli [11:06 PM]
Ah, that would explain. I thought it was fixed. And usually it seems it is

omei [11:07 PM]
I had assumed it was fixed, too.  But Johan doesn't alway keep inquisitive players in mind when he publishes the results.

eli [11:07 PM]
Ok, that answers it

Mirror structure

Basically what I am wondering about is if the structure of the ON switch tell on the structure of the OFF switch or vice versa.

I have already seen this work for the input labs, where the ON switch tends to prefer opposite order of the A and B inputs to the OFF switch. (Logic gates + link)

I think this is worth looking out for. This may help us solve in new and different ways.

Exclusion - Theophylline B

Score 100, fc 29.26


Same State - Theophylline B

Score 90, fc 13.07


Now the design with the highest score in this round isn't the highest scoring one. This one is.

While it is not a structural inversal as the above, there are still some structural similaries between the two states.

Score 100, fc 34,97


The same is the case in one of its siblings. Here there isn't a mirroring though.

Score 99, fc 24.55


Score 95, fc 20.62


It is kind of a different structural pattern to what has already occoured a lot in riboswitches, with the two states taking sort of a crossing movement. Here the folding lines are just a bit different.

This kind of structural mirroring or repeat don't go for all the best scoring designs. But this may have potential.

So look out for structural patterns that may help tip us off on what puzzle design is more likely to work.
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes
A sequence signiture of lower scoring designs

I have taken notice that longer stretches of G and U seems to badly effect the score.

So far I have gone through all the theophylline/MS2 lab data and my suspision has only been confirmed. Designs with longer stretches of GU's disproportionally ends up with lower scores + raised fold change error.

However there are some good scoring designs that do also hold a stretch of G and U, but I think position matters. It may be worse if it is close to or inside the switch elements. Also not all combinations of G's and U's seem equally bad.

I suspect this is why that we have gotten less good results on the Kissing loop labs, as the Kissing Loop itself holds this G and U pattern. In other words, I think we will get better results with kissing loops, if we just pick one with another - and more switch like sequence.

Discussion with Omei from 27 Feb 2018

eli [12:13 AM]
In most of the labs it looks like MS2 performs better than KL
But then again, KL doesn't have MS2 complementary bases, as did the FMN aptamers and other of the aptamers
So less luck for direct pairing up for shutoff

omei [12:16 AM]
It seems like the KL locked few enough bases that additional stem was required.  Why couldn't we create whatever complementary bases we wanted as part of that stem? (edited)

eli [12:21 AM]
Ok, we could and we did.

eli [12:21 AM]
uploaded this image: image.png

eli [12:21 AM]
But there is something else different about this switching hairpin than MS2
It holds excess G and U.
I have regularly seen excess G and U in designs that don't switch well.
The kissing loop labs generally don't reach high fold change
I'm not certain they are to blame, just wondering

omei [12:25 AM]
Do you consider the design you pasted to have excess G and U?

FWIW, Joanne is inclined to believe that the KL aptamer doesn't bind as well as was described in the paper he got it from, at least not under our experimental conditions.

eli [12:30 AM]
No, just the kissing loop itself.
I recall earlier labs where I tried a GU rich turnoff sequence instead of a CU rich one. And it tended to fail miserable.
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes
If this G and U sequence excess being bad is a general pattern in relation to switches, it should probably be taken into account when RNA inputs are chosen.
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes
Shorter G and U stretches don't spark as much trouble as the longer ones. I have regularly seen stretches of up till 4 bases work really well and even be needed.