Blueprint of the MS2 and FMN Riboswitch

  • 4
  • Article
  • Updated 2 years ago
  • (Edited)

Omei and I have been discussing RNA. I have been enjoying it a lot. One thing we have been friendly disagreeing on. Basically Omei got me set off taking a stand on what I think I know so far. So I’m bringing a part of the discussion with some additions.

Omei: I should say that I don't share your confidence that for any lab there will be a unique good way to design a switch.  Here's my rationale.

Taking Exclusion 3 as an example, we get to chose one of 4 nucleotides for 42 bases.  That gives us 4^24, or about 2 x 10^25 possibilities. If there's essentially only one way to solve the puzzle, how many variations of that one way might there be?  Lets say there are a (US) billion = 10^9.  That would imply that only one over 10^16 of possible submissions would be a "winner".  Do you really think we understand RNA so well that we could find a winner at those odds with only 3000 tries?

Actually this is exactly what I think.

Here is why. I basically keep seeing ground structure with the same elements pop up across designs in a lab and also across labs. MS2, FMN, MS2/aptamer turnoff and static stems at particular spots. These elements have preferred position. They want similar distance. Only a certain range is acceptable to each. But it all adds up to a puzzle that wants to get played in a very particular order and structure.

Notice how much these different labs resembles each other in basic structure and mirroring. Notice that the only difference for the static stems (blue) is an open ended stem versus a hairpin loop stem.

Structural drawings of ms2 fmn switchesjpg

I think there is a really small range for a good solve. I think there maybe are like 100-200 really good solves with good score and fold change for each of the 3 mentioned labs. And that the main part of these will be very close in sequence space and overall structure. Of cause with more change allowed at static stems than elsewhere.

Nature Versus Simulation

On the amount of possible solves. Lets think about static designs. While they can tolerate far more solve variance than switches, there really isn't such a huge range for good solves either. Plus usually we hit a winner in 1 or 2 rounds. Why? Because I think nature has a much more limited list of what it fancies than simulation. Nature is far more consistent and "dull" compared to how it is possible solving in puzzles.

The puzzle simulation give us far too long elastic as they are letting us get away with Christmas trees, Cub Scouts, Optical Illusions and wildly uneven energy distribution, compared to nature. Plus the many other rules that sets borders for what will be a legal solve. We can probably get the simulation eat 10-20 times (wild guess) as many potential solves than lab will accept. Those potential “solves” will be very abundant.

Also the puzzle simulation give far too long elastic for making very hard puzzles. Given the choice between one of Wawan’s puzzles against a lab puzzle, I personally will choose lab anytime, as nature is far more forgiving. :)

I suspect that most of Wawan’s puzzles will have a rather limited set of possible solves. Plus I found it doubtful that main part of these puzzles will ever work in lab, let alone nature.

Basically nature is far more forgiving than both a Wawan puzzle that allow some few solves but less forgiving than a more sloppy puzzle that allow for a million different solves.

Nature is far more rigid and thus predictive which isn’t too bad as we can totally play this to our advantage. By following the basic rules for what a lab wants, be it for structure, base distribution or sequence for a particular element, we can simply beforehand rule out a multitude of solves that very likely will not solve in lab, despite being allowed by simulation. Plus I even suspect we can in the same go also rule out many puzzles as likely lab solves, if the simulations take far too long time solving or can’t solve solve them at all. Imagining they are like Wawan and the many other player puzzles with structures and/or sequences that are unfit for lab.

I think is switches are quite similar to static puzzles in regard with a limited set of likely solves. Except there is a much more narrow window of solving opportunities. The preferences of each switch element, MS2, FMN, MS2 and FMN turnoff, static stems etc, each put a limit on each other and allowing for only very small subset of really solves.

I love the narrow twilight zone for RNA folding, between the way too strict and far too sloppy for nature.

Same Switch Blueprint Across Labs

Here is how I got to the image I showed at the beginning of the post.

A while back I did some drawings on switches from memory. I drew this one below of what I thought the NG labs really wanted. I think the tails are unnecessary, I just drew them to demonstrate that I kicked out what I think is excess bases from the main design.

NG round 2 predictions

Cut and glue doodles2jpg

I later made this notebook drawing.

Rough draft2jpg

I had earlier realized that for some of my earlier lab drawings based on actual solves, some of the NG solves would overlap with the Exclusion and Same state solves, if just rotated.

I decided to regroup the NG drawing guesses, these so they fit those my notebook page. I redrew both my lab sets drawings.

Riboswitch on a chip + NG 2 (Notebook)

Riboswitch on a chip structural drawingsjpg

NG labs + Same State 1

Ng labs structural drawingsjpg

Now both these drawings are a bit stylized, as the NG one was a guess on the Round 2 NG results and I did the notebook one based on memory. But I think they show better the relation between the structures in the different RNA’s, but also the similarities for structure even between turn on and turnoff labs. Not every winner look exactly like this, but there are several close ones and I think they are aiming towards this pattern.

Some of the exclusion designs have their legs more kicking to the side. Exclusion 3 designs in particular, as their MS2 G’s comes quite early in the MS2, as it is positioned in relation to the aptamer. These G’s are by far the preferred target for turning off the FMN and MS2 in Exclusion labs. But what I find interesting is that there still is so much overlap in structure between as different labs as turnoff and turn on labs, and across the lab designs in each of these in general.

Here are two real life winners from a turnoff and a turnon lab. (State 2)


On the structurally similar but different labs on the NG labs that structurally resembled the Riboswitch on a chip labs:

Does it matter if a static stem in a lab is a neck or a hairpin loop stem? Probably. Does it matter which way the FMN aptamer turn in relation to the puzzle? Most likely. While the FMN aptamer works both ways, it seems directional and I think one way works better than the other.

There will be variations - posed by limitations of how the switching elements are placed (FMN and MS2). Not each positioning of elements works equally well. The FMN’s can take turn on being target by a turnoff sequence (Targeting FMN1 overall works best) and the MS2 can have its turnoff on either side (Right side works best), MS2 can be before, between or after the FMN sequences. (Between works best). Generally one option is better than another.

So I think the one switch that positions all of its elements so they are most possible happy each, is going to be the best one.

Riboswitch on a Chip Perspective

I simply think when the length of the sequence allows for us to make a partial moving switch, for making static stems in relation to the switch elements, then we can hit a legal lab solve pretty fast. There also also needs to not have too far or too short distance between the switching elements (FMN and MS2). Both of which Same state 2, Exclusion 3 and Exclusion 2 fulfills perfectly.

For Exclusion 1 especially and Exclusion 4 I think we have already found most of the best possible solves (when it comes to partial moving switches) and it is only few details we can change. Not all switch structures can become fully happy and content switches, they have too many structural hindrances build into ever become good switches. The positioning of their switching elements makes all the difference in the world in relation to if they are solvable or not.

For Exclusion 6 in particular but also Exclusion 5, I think there are more structural ways to solve, but even these labs have their own set of rules to follow. (At bottom of this post) Just it does not follow the overall general pattern as that I see for the switches that allows for good partial solves. Yes, one can a rare time solve with a full or fuller moving switch style, I just think it is far harder hitting on it.

I think there are only a very narrow range in structural and sequence space that will actually solve partial moving switches but luckily it is reusable across labs. I could get my deletion of the MS2 gates to work in other labs than the one where I first realized that deleting it might be beneficial.

So I remain very optimistic on behalf of the partial moving switches and there being a collective overall pattern for a solve. I see the shade of the same basic structure behind every lab. Almost everything I see points towards it.

Perspective on Riboswitches in General

Ronald Breaker, riboswitch pioneer, was on to this a long time ago, (I wrote an intro to him and his work here.) with reusability of positioning of switch elements. (Riboswitches and the RNA world, Ronald R. Breaker, 2006, Chapter 4)

Basically he and his team got a particular ribozyme (Hammerhead - self cleaving switch) working together with an aptamer. And the really beautiful part was that they could interchange aptamers and still get the ribozyme working.  

The riboswitch collection in Figure A, page 92, called AR5, kind of looks structurally related to our Exclusion NG 3.

So now I wonder how our riboswitches are related to those ribozymes.

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes
  • Riboswitch T-rex

Posted 4 years ago

  • 4
Photo of rhiju

rhiju, Researcher

  • 404 Posts
  • 123 Reply Likes
Eli, the equivalence of successful blueprints -- even across puzzles that have distinct ordering of MS2 and FMN elements –  is a very very cool insight. Those blueprints are the same up to 'circular permutation', in the parlance of the field.

One thing that would help me to understand better is to also show a gallery of schematics of blueprints that we *don't* think will lead to experimental success, perhaps with some explanation of why.

And after that, we could use your and players' help to test the generality of these ideas beyond FMN & MS2, for example in new same-state and exclusion labs that use aptamers for other small molecules (e.g., theophylline, ATP instead of FMN) or that bind different proteins (e.g., L7ae instead of MS2) or miRNAs. The question will be: could we define a set of blueprints that should give high rate of success, and blueprints that give low rate of success? The cleanest test would be to actually have eternabot design the sequences, given the distinct blueprints -- johan & wuami are both excited and ready to do this!
Photo of Brourd


  • 461 Posts
  • 84 Reply Likes
It should be possible to test some of these hypotheses, such as

Does it matter which way the FMN aptamer turn in relation to the puzzle?

by taking the 5' AGGAUAU 3' - 5' AGAAGG 3' FMN aptamer and rotating the structure with respect to the strand, to 5' AGAAGG 3' - 5' AGGAUAU 3'.

Part of the reason I agree with Omei with respect to this matter isn't so much the capabilities of players to be able to accomplish the task that is given to them, but rather that players are not providing evidence that either corroborates or falsifies the hypotheses they propose. As an example of me committing this sin on my own part, we can compare Exclusion 4 and Brourd's Mod of Exclusion 4.

The objective of Brourd's Mod of Exclusion 4 was to determine if increasing the length of the 5' and 3' ends of the strand would have an effect on player scores overall. Originally, it was anticipated that players may take advantage of the increased sequence space to design more diverse secondary structures, but this did not happen as often as hoped.

The average design score for Exclusion 4 was 59.3, and a median score of 61.93.
The average design score for Brourd's mod of Exclusion 4 was 60.56, and a median score of 62.11

From this result, we can probably make a rudimentary conclusion:

With an increase in the length of the dangling ends, player performance did not increase by a significant factor, and therefore the length of the dangling ends most likely does not directly factor into the function of Eterna riboswitches, unless there is an extreme factor involved.

We as players have the potential to further build upon these and other hypotheses and conclusions in order to develop a unified theory of RNA riboswitch design, but I think this system needs more than just sequence design. It is my opinion that the ability for players to create secondary structures to test hypotheses is just as important.

To elaborate on what I mean, one hypothesis I had from that previous conclusion was that maybe the addition of a helix "tying" the dangling ends together would have a positive effect on score. The way to compare this would be two designs with two distinct conditions. One where the dangling ends can fold into a helix ~6 base pairs in length, and a second design identical to the first except for that helix, which would be designed to be unpaired.

Unfortunately, I don't have the time to design an experiment like this, and the apathy I see from players for the design of experiments is much higher compared to them just solving a puzzle that has been provided to them.

Anyway, there are other hypotheses that I've had for various sequence and structure conditions, but rarely do I feel like I can justify putting work into proving or disproving them given the limited number of design slots and the amount of time to set any one of these experiments up.
Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

Pointers for the EternaBot

Hi Rhiju, Wuami and Johan!

I have made some pointers for you guys and EternaBot.

Here are some arrangements of functional elements that I have seen tends to give bad results both for EternaBot and for players.

No static stem at the aptamer end opposite the MS2

Things that is bad for making a successful partial moving design is if the FMN aptamer is not properly stabilized in the end that is opposite the aptamer end there the MS2 is.

Here is an example from EternaBot. The aptamer is closed at one end but only with a short weak neck. So weak that it is broken open in the other state. The tail ends should pair up also, to ensure that the neck holds and the aptamer has one stable end.

Exclusion 3, Score 12%

A good partial moving switch mostly tend to have the end of the aptamer locked. Noticed the greatly reduced switching area. The fewer bases/base pairs we need to have switching, the easier it is to do it right.

Score 100%

Why: Why are two FMN sequences flying around apart from each other in space a problem? Because it is far harder making the two FMN sets meet at the right spot at the same time, when 2 stems have to move but also become stable around the aptamer sequences. To make the FMN working, both its sequences need to be close in space. By locking them at one end and doing so with a good amount of base pairs, we secures that they always stay close in 3D space, so they have a fair chance to get turned on when FMN is around.

Has both end of aptamer switching worked in some labs? Yes. It has worked in half open switches like Ex 1 and Ex 4, where both stems around the aptamer has been moving, something Omei noticed (Internal loops and switching). The reason I think this helped, was because it gave a switch that had only a very small sets of options to switch, some extra moving space. In case of the hard labs it is simply needed for allowing the switch to go full moving, as it is the only way to make a solve. I have seen some rare cases in otherwise partial moving designs, where the stretch closest to the end aptamer is more flexible and perhaps opening.  

So if the neck (or hairpin stem if in a NG lab) is broken in an otherwise successful (partial moving) switch design, then scores should go a lot worse. And adding a neck should help - if that lab is not one of the ones with too short tails to make a neck (Ex 1 and Ex 4) or one that due to having its switching elements too far apart, demands a fuller moving switch Ex 5 and Ex 6)

Background articles

See section called Tails

See section called Static aptamer end

Switching parts too far apart or too close

Hard switches - have their switching parts placed too far apart for them to get in reach in an easy manner and thus needs the switch to go full moving to even have a chance to solve.

Why: Any stem formed between MS2 and FMN has to get broken to make the switch move, and only makes a switch harder to get moving. So if both the aptamer gate is too long, the MS2 gate is too long or there are even additional stems to be broken, getting the FMN and the MS2 to hug each other is going to be a lot harder.

Now the winner and high scorers of Exclusion 6 do have some conserved pattern to them, but how to hit on it in the first place? I suspect that each full moving switch are going to have its own individual blueprint. But not as shared as for partial moving switches, which blueprint seems reusable no matter how it is turned.

How to make hard labs

Exclusion labs: Make distance between the FMN and the MS2

Same state labs: Put MS2 right next to FMN

Distance between MS2 and FMN makes solving in Exclusion labs harder. A distance of 4 bases however is seemingly harder than an even greater distance or too far apart with no static stem in between.

Exclusion 3

Same state labs gets harder to solve if the MS2 gets too close to the FMN (2 base distance min). If the MS2 gets too far apart from the FMN, it can be brought close again by making an abbreviation with a static stem of the excess bases.

The MS2 has favorite distance to both sides of the loop between itself and the FMN sequences, in the Same State labs.

If there are not options of making static stems to abbreviate long base sections between switching elements, then the switch is going to get hard to get moving.

Recipe for making a hard lab :)

Now to something funny on hard labs like Exclusion 5 in particular and Exclusion 6 to some degree. Plus basically any switch lab that doesn’t have its switching part close enough together for them to have a good shot of doing some hugging. These labs won’t follow the same blueprint that the partial switches do. They will have each their different blueprints instead.

List of things that don’t work well in pressured switch labs.

  • Multiloops

  • Lack of GU’s

  • Static end of the aptamer

  • Static stems

  • Just one shared MS2/FMN turnoff sequence

Why: This kind of lab despise any kind of element that makes the design more overall stable, well except perhaps multiloops, that tends towards making things more unstable. However they usually also comes with a stabilizing static stem. :) The harder the puzzle, the more counterproductive any static stem will be.

Background article

Pressured Switch Labs

Advantage of partial moving switches

Working from the image above only 58 bases in total needs to be switching. Of which 19 (MS2) and 13 (FMN) are leaving only a minority of 26 bases to be mutated. And not all of those are of equal importance and some of them even have preset preferences. Where as a full moving switch and hard lab will literally have all of its bases moving. Making numbers of potential solves go way up and being much harder to find.

Back to partial switches. The static stem in the switching area needs to be solved in single state manner and mainly the closing base pair in the multiloop and next pair after are of bigger importance. So this also subtract from the space of bases that really need to change. Likewise the turnoff sequence for Exclusion labs, is overall pyrimidine and pretty much conserved with smaller variances between the different labs. Similar the static stem at the aptamer end can be solved as a regular static stem and won’t need much attention as long as it has some good base pairs next to the aptamer.

Leaving mostly the position of the static stem and the loop bases around it to be mutated. As also the aptamer gate bases are pretty much determined by the MS2 sequence. At least for the partial moving switches. Something similar takes place in the Same State puzzles, although with a bigger room for mutation at the aptamer gate and the loop area around the MS2 and the static stem in the switching area.

For many switches this space for change will be enough to make a proper switch. And for some switches that can’t get entropy high enough in such a small space, some more bases may need to be added for allowing the switch. Or perhaps allowing the aptamer to move just a bit at the “static end” while still being securely attached to a longer static stem.

For me my favorite game till now has been to back all the entropy up into a small corner or rather a circle bubble of switching elements - held between two static stems.

RNA Slingshot

Rhiju, you got me inspired with that circular permutation thing. :)

I have for a long time been imagining the partial moving switches, kind of like two sticks (the two static stems) with 2 elastics folding and curling between them. Depending on how I hold the sticks in relation to each other, the two different states take place between them. The long elastic is like 3/3 to one 1/3 of the short. (This length relation is typical for both Same State 2 and Exclusion 2 and 3 partial moving switches.)

Same State

The long elastic always holding the MS2 and an (aptamer gate) complement to the short elastic (aptamer gate). The longer the rubber band, the more it likes to curl with itself. The shorter the rubber band, the less curled.

Same State 2, 100%


The long elastic always holds the MS2 and an MS2/aptamer turnoff (pyrimidine stretch) that is complementary to the FMN at the little elastic. That is if it the design has no MS2 gate. If it has a MS2 gate, then it seems as ifthe really long elastic curling up with itself. Leaving the little elastic to curl with itself.

Exclusion 3 - With MS2 gate 98%

Exclusion 3 - No MS2 gate 100%

Adding MS2 gate

After this, I really would like to see some heavy EternaBot computing on the topic to make a MS2 gate or not. I think in many cases MS2 gates will hurt an Exclusion partial moving switch or put another way, while we in some labs can make winners with them, I think we may be able to do even better without them. While I still believe we need MS2 gates for microRNA labs.

Basically only the MS2/aptamer turnoff sequence needs to change to change target back and forth between FMN1 and FMN2, which will decide weather or not a MS2 gate is formed.

Background article
Change aim for the opposite FMN
The curious case of the missing MS2 gate

Breaking the Exclusion labs

How: Easiest way of breaking the Exclusion labs solves and their switching ability, would be to ban the pyrimidine MS2/aptamer turnoff, that is most often found just right before or after the MS2 sequence.

Why: Same State labs tends to like direct hugging between the switching elements (MS2 and FMN (sometimes including some of the aptamer gate), while the Exclusion labs prefers a middleman sequence (mainly pyrimidine) that are complementary to both FMN and MS2.

The pyrimidine MS2/aptamer turnoff is strongly preserved across Exclusion labs, even for the labs that can’t do a partial moving switch. So disrupting the pyrimidine stretch, should disrupt the switching.

No static stem

Exclude the static stem from partial moving switches.

Idea: What I wish to test is if a MS2/FMN riboswitch will work just as well if it doesn't have enough bases to form a static stem in the switching area or if that static stem is just an artifact of excess bases.

Why: Ever since I realized that a static stem popped up in a huge part of the winners, I have been wondering, if the static stem was really needed. Although Brourd’s mod of Exclusion 4 showed that a winner can be made without, I still think the static stem actually do has a positive effect on the switching. Like by resulting in a multiloop, making things a bit more unstable = better switching.

How: I was interested in testing it in Brourd’s mod of Exclusion 4, as it already has all the excess bases gone. But besides having an extra base pair between MS2 and FMN it also has the aptamer open at the opposite than Exclusion 2, Exclusion 3 and Same State 2. However I think a test can be done by running the main structure and sequences of the Exclusion 2, 3 and Same State 2 lab winners and simply remove those “excess” bases that tends to become static stem in some of the winners. And then run mutation in the switching area. In those 3 labs we already have good results for comparison.

Images of cutting out the static stem bases in Exclusion 2 and Same State 2

Same State 2

MS2 not held from both sides

MS2 is missing backbone connection from one side, so it is not connected up with both FMN sequences. (Happens in Exclusion 1 and Exclusion 4)


Why: This is a problem, especially because I think the MS2 is very strong. The naturally occurring open ended switches I have seen in RFAM, didn't have a MS2 they needed to shut off. It takes a good deal of force pulling a MS2 apart. Catching MS2 from both sides, when it needs to get turned off, I think helps. It it tends to need to get hold from both ends to want to stay open.

Background article

Exclusion 1 - Round 4

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

RNA Slingshot continued

I continued my thoughts on the RNA slingshot and RNA stems holding a RNA circle. I did some drawings.

Exclusion 3/Same State 2 type

Bands lengthjpg

Notice how each of the static stem is parallel to the folding up in either state.

Same state circlejpg

Here it gets really interesting. Notice how the circle is dragged askew, when there is a MS2 gate forming. The switch element attractions are no longer parallel with the static stems.

Exclusion 3 circlejpg

Also notice that both Exclusion 3 and 2 in particular are more askew, even in their version without MS2 gate. The pull is less askew in the Same State 2 lab.
Exclusion 2 circlejpg

Here I have highlighted the actual switching and binding segments.

Two and twojpg

The elements are basically placed so they only fit two and two.

In all cases these switch folds are doing nearest neighbor strand partnering. Ok, the aptamer gate pairing with the MS2 beginning sequence - is a long distance cross over. But since the locked neck and static end of the one end of the aptamer is already in place, the crossing over is already done far easier, as the connection has already been made. Similar for the static stem in the switching area. It is more like a next neighbor strand pairing they have, because there is already a static stem formed and just an internal loop between the static stem and the close by pairing up.

This image probably illustrates best. There are really only two possible pair ups for that top right corner quadrant. Exclusion 3 1jpg

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

RNA folding - Origami style

I was thinking about Chris Cunningham's awesome post on RNA folding with paper and glue and decided that my circular RNA could fit neatly on a origami fortune teller game.

Exclusion 3 type

RNA fortune state tellerjpg

Full view.

There really only is two states.

Second state.

Ok, actually there is a third state in between. The one where the MS2 gate forms - somewhere between state 1 and 2. Holding it with a paper clip. :)

However the MS2 gate takes far too much energy and trouble creating. Plus switching to it is real hard. There isn’t anything like the natural flipping back and forth between state 1 and state 2 in the RNA state fortune teller game.

MS2 gates aren’t energetic effective. MS2 gates needs to go. RNA origami proof over. ;)

Afterwards I found an older paper on circular permutations and RNA. (Encoding folding paths of RNA switches.) When I saw the image in Fig 6 I understood. I realized that this is exactly what you have been talking about, Rhiju. :)

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

Different aptamers, same blueprint

Continuing in the circular permutation line of thought and addressing your question from earlier:

Rhiju: And after that, we could use your and players' help to test the generality of these ideas beyond FMN & MS2, for example in new same-state and exclusion labs that use aptamers for other small molecules (e.g., theophylline, ATP instead of FMN) or that bind different proteins (e.g., L7ae instead of MS2) or miRNAs. The question will be: could we define a set of blueprints that should give high rate of success, and blueprints that give low rate of success?

Even better. I think we can use exactly the same circular permutation blueprint with different aptamers, with only little variation needed. As long a few ground rules are followed.

Important question to ask for aptamer complementary.

I think that the MS2/FMN blueprint will even hold for other aptamers interchanged at same position, if just these aptamers/switching element complies to a few demands.

  • Do they both holds stretches that are directly complementary - as MS2 and FMN do in Same state - so they can do direct hugging?)

  • Do they both have have identical sequence stretches that can be used for a middle man sequence (kernel attractor) as FMN and MS2 do in Exclusion labs?

  • Do they have magnet segments?

The ATP aptamer resembles the FMN very much in sequence with lots of G’s and some A’s, except it is being stronger. (The ATP aptamer comes in a single and twin aptamer version - the twin actually holds two ATP molecules side by side, with the same loop. :) ) It should be possible using it as FMN, with both a direct pair up with MS2 or L7Ae and with a middle man sequence.

Not all aptamers will fit each other equally well. But it should be possible matching up aptamers, just by watching their complementary, shared sequences and strength of their elements. So one could set them to switch fast by them being balanced in strength in relation to each other. Or one could set them up to be slower if that somehow was useful at times too.

The TEP has a partly complementary stretch to MS2 and FMN.

One common FMN turnoff sequence

TEP aptamer sequence found in 4 x 4 TEP riboswitch 5’ lab

Here is an example with the TEP sequence - as complementary and capable of turning off both the MS2 and FMN aptamer.

TEP has magnet segments. Twin G’s like the FMN.

Although I should add that FMN is particular suited to flip full way around the circle permutation, no matter where the FMN is placed. Due to its mirror nature, with two twin G's in each side of the FMN. This means that it will behave somewhat the same, no matter from which end and from which side you approach it. TEP won't do the same. It will be far more suited for a specific aiming for a specific stretch. And for one side it will want something but if aiming for the other, it will want something very different. One can't just flip a few bases around and then aim for the opposite aptamer sequence.

Some aptamers will have more limitations than others. Its a matter of how much they resemble what they replace. The TEP aptamer will only have half the amount of the possible blueprints I showed for FMN. One can't get the mirroring of the blueprint around the vertical axis with TEP, as one can with the FMN aptamer. Because the TEP does not contain the mirror pattern that is in the FMN. (Its sequences are close to mirror patterns) So there will be fewer of the blueprints that I showed for the FMN that will work. At least if one expect to reuse the usual pyrimidine MS2/FMN turnoff sequence. Or rather, the mirror version will need a different turnoff sequence. :)

Do L7Ae have similar sequence stretches to MS2? In particular do it have identical stretches? Do it has strong G and C magnet segments like MS2? Check, check, check!

Magnet sequences highlighted

L7Ae versus MS2jpg

L7Ae sequence found in fig 1B. Thx for the paper on L7Ae, Rhiju!

I almost can't believe that the L7Ae doesn't even have a single U. Thats odd. :)

The long stretch of C’s and G’s comes early which means they should be easier to get to, to pry the L7Ae open for disruption. And they come on the same side as the MS2, which gives hope for interchangeability.

L7Ae even seems to have the ability to be hold from both sides, just like the FMN aptamer. (Fig 1A)

Blueprint reuse - some reductions and changes will happen

I see absolutely no reason why we can’t interchange the aptamers you mention for the FMN. Similar I see no reason why we can’t interchange L7Ae for MS2. Not each set will fit equally well, and each match up won't necessarily give allow a full set of usable blueprints, but I think it will generally be doable, getting to a switch that works.

If the aptamer replacing FMN doesn't have mirror symmetry, then the mirror blueprints will need a different turnoff sequence, than the other blueprints, for those blueprints for the side of the aptamer that do have twin a G stretch.

So basically it do matter if the magnet segments sits at the same side in the aptamer one wishes to interchange and if the aptamer itself has sequence symmetry. But if this is taken into account, then it is

Perspective on interchanging aptamers

I have already been doing some thinking of FMN versus TEP and some of the other aptamers.

I earlier found that some aptamers had a strange similarity in sequence. They hold similar sequences and same types of bases. Like longer stretches of purines or pyrimidines. I found such sequence endemic in switches. I simply think there are specific base arrangements that are especially good for switching.

While I knew that MS2 and FMN used a middleman sequence (Exclusion type) and that they could do direct hugging by complementarity also (Same State type), Omei got me thinking about MS2 and FMN in a new way. Which lead to me speculating about further implications for other aptamer pairs. I see sequence patterns going through what makes a good aptamer.

That whole post is basically about that. But I will specially highlight this section as central:

The FMN piece inside the MS2 hairpin

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

Does microRNA fit the MS2/FMN blueprint?

I have been doing some thinking about the re-usability of the MS2/FMN blueprint in relation to microRNA.

Rhiju: And after that, we could use your and players' help to test the generality of these ideas beyond FMN & MS2, for example in new same-state and exclusion labs that use aptamers for other small molecules (e.g., theophylline, ATP instead of FMN) or that bind different proteins (e.g., L7ae instead of MS2) or miRNAs.

First I have taken a shot Exclusion 2, 3 and Same State 2 high scorers and highlighted their magnets segments. Then I have drawn images based on them to highlight their nature.

Second I have drawn the schematics different to highlight the magnet segments as their blueprint position matters also.

For the MS2/FMN blueprints there are basically three things going on. First structure (T shape), then sequence (word change gaming) and then attraction (magnet segments).  

  • A T shape structure

  • A pattern of base stretches pairing up

  • A pattern of magnets

These two latter do not always overlap.

The word change game

The Same State 2 blueprint is the most flexible, since it leaves the aptamer gate a little less locked with the word change gaming. Where in the Exclusion labs, the MS2 needs to be right next to the FMN, which leaves the aptamer gate pretty much locked in sequence. And the MS2 turnoff sequence is rather locked too, leaving little room for changes. Whereas Same State don’t need the rigid pyrimidine turnoff, but can shift things a bit more.

Pattern of strands matching more than one partner. Basically it’s a word change game, not unlike the one going on in the microRNA labs.

Example from the microRNA lab:


Example with MS2/FMN blueprint

Exclusion word change gamejpg

Word change game - Big locked sections

Just like the microRNA’s and the MS2 are the fixed sequence for the microRNA labs, so is FMN and MS2 locked. Except that there is even more locked sequences. The exclusion lab is the most locked.

Since the MS2 needs to be next to one of the FMN’s in exclusion labs, the one aptamer gate that is covered by MS2 sequence is thus locked and its opposite partner is pretty much determined by complementary.

Secondly the MS2 G’s are by far the most favored sequence for turning off MS2. And thus this section is also locked - to be both complementary to the MS2 G’s and one of the FMN twin G’s.

So the word change game part of the puzzle has very little change allowed.

Whereas in the Same State 2 lab, the aptamer gate is a little less locked. It should be complementary to a stem in the MS2 in some way and then with itself.

I called it word change game. But due to the nature of the switches, I could call it pyrimidine and purine pairing game. :)

Illustration of blueprints for magnet segments

Here are the main blueprints till now, as ranked highest, left to right, and top till bottom. This image is a sum up of of my magnet section images shown below.

Ranked magnet blueprints leftright topbottomjpg

Pairing strands + magnet sections

Magnet segments highlighted with colors. Unused G magnet segments/purine stretches are shown as thin red lines. Notice how most of the switching area is either pyrimidine stretches and purine stretches (primarily the latter), for the exclusion labs without MS2 gates.

Exclusion 3

Score 100%, with no MS2 gate

Score 100%, with MS2 gate

Exclusion 2

Score 97%, without MS2 gate


Same State 2

100% (Single magnet system)

96% (Double magnet system)

One thing to notice, is that designs with MS2 gates for Exclusion 2 and 3 have their 4 magnet segments in the big portion of the elastic. Whereas the ones without MS2 gates, have one magnet in the small portion of the elastic and 3 magnets in the long one.

Drawings of magnet blueprint segments

What I wish to highlight, is that if a aptamer, be it loop (ATP or TEP) or hairpin (L7Ae) should be interchangeable with FMN and MS2, then it should fit the magnet blueprint also. Not just the structural shape it replaces.

The drawings are equivalent to the screenshots of lab winners shown above. The bottom of the drawings are shown as circle bubbles instead with magnet segments highlighted and arrows for pairings.

Exclusion 2 with ms2 gatejpg

Exclusion without MS2 gatejpg

Exclusion 3 with ms2 gatejpg

Exclusion 3 with no ms2 gatejpg

Same State 2 - single magnet systemjpg

Same State 2 - double magnet systemjpg

Not all of the magnets are involved in forming stems in both states. Eg the FMN magnets are most not in use both of them, and in the turned on state, they are forming a loop holding a FMN molecule and thus not being involved in base pairing. The FMN sequence is bound as a stem in in the state where one of them is turned off.

However there are multiple possible setup possible, which is why I’m generally positive about interchange of element. As one can choose the set up where the magnet in the switch element one wishes to change in, fits the magnet pattern of one of the working forms.

Aptamer inputsjpg

This is basically why I say L7Ae can replace MS2 as it has the same magnets as the MS2 and positioned similar - although it also has an a extra set. Similar the ATP aptamer has similar purine base stretches to the FMN and as such should be interchangeable. The TEP aptamer has one twin G stretch so it may replace the FMN or the MS2 as it also has both a C and a G magnet stretch.

Can these blueprints be used with microRNA’s also?

That will to a big degree depend on the microRNA. And if its magnet segments fit too. Also it should be prepared for word gaming with the aptamer it is getting partnered up with. If it isn’t somehow complementary (Same State blueprint potential) or has identical stretches with the aptamer (Exclusion blueprint potential) then it may get a hard time.  

There are also more profound issues for the microRNA’s in relation to getting them pair up inside of partial moving switch.

MicroRNA may represent a challenge for changing directly with MS2 in the MS2/FMN blueprint, as microRNA have some special properties.

I ran a quick and dirty experiment on entropy of a bunch of microRNA’s and they don't at all have entropy for something that would end up like a hairpin stem, albeit a stable of them. Those I ran through Vienna, has high entropy. Although this is similar for MS2.

From my meeting with microRNA’s to now I think they are different from regular hairpins. From what I have seen till now, they are not too happy about forming hairpins with themselves.

Also the microRNA's don’t always have magnet segments, also they don’t necessarily have both kind of magnets, like the MS2 does.

MicroRNA have a tendency for being skewed in base distribution. Eg having way more C’s and very few G’s. Or opposite. Meaning they may have double C magnets and as such can not directly fill at the MS2 spot as the MS2 has both a G and a C magnet system. It may still be possible doing this, we just have to change the blueprint so it fits to microRNA’s.

Or in the case with TB B, that has mostly C’s, so it kind of have two green C magnets. Which makes it a bad replacement for the MS2 that has both a G and C magnet stretch.

The TB B having most C’s will require a G’s turnoff sequence. If the FMN G's can be used for that fine. But if not, there will need to be a separate G’s turnoff sequence. Till now we haven’t had much success with these. At least not for turning off MS2, despite it having a fine C stretch that hypothetically did offer the opportunity. I have tried that, but I haven’t gotten it to work. So the Exclusion blueprint will be bad. So there is the little less locked Same state pattern. But still there may be problems. As usually the MS2 likes to pair up with both its ends to each of the aptamer gate strands. And both can't be heavy in G's. But I think the Same state blueprint will be best for an attempt.

Switches also seems to be extremely fond of longer pyrimidine and purine stretches, something which microRNA’s have not necessarily evolved to love. They need to bind to a messenger RNA which is something very different to a switch. Whereas I keep seeing the pyrimidine and purine pattern in the switching parts of riboswitch. Plus as I noted in the post above. The reason why MS2 and the FMN can work together is that they have similar stretches. But they comes from each their organism,  Bacteria and virus, and has not been together in nature in the same organism working together, as far as I’m aware. But yet they fit, like hand to glove (tight glove, but still), due to identical sequence and complementary stretches. So I suspect that switching elements have something identical about them. Something which helps a switch get moving.

Why may the microRNA’s be hard to fit inside a partial moving switch?

First the microRNA seems to be fond of doing a full or almost full pair up.

Second, microRNA’s seem to need a longer complementary dangle to attach to a RNA design. Such a dangle isn’t happy to stay available in a multiloop, an internal loop or an end loop. It is far the happiest at the dangling tail end of the RNA. Last rather, but first also. Plus if neither is possible, then gap bases will do. This is why I think it may going to be harder to get a microRNA to attach at the inside of a closed RNA design. Also the switch labs have shown to not like having a high single base to base pair ratio. Also the partial moving switch lab winners so far do not want too big loops. The best partial moving switch designs mostly have smaller multiloops.

I’m not saying it can’t ever be done, just that I foresee that there are going to be some issues. And that microRNA may take some more fitting than natural occurring switch elements.

With different aptamers it might be possible to create a blueprint that fits to a certain microRNA.

Perspective on microRNA labs and their blueprints

However our more open microRNA labs, still do carry part of the blueprint anyway.

The winners in the turnoff variant 1 and 2 lab, the microRNA pairing op forms a multiloop bubble with the RNA design sequence. This switch still wants its multiloop. :)

The MS2 still needs a turnoff. It is still favored at the right side of the MS2. A turnoff for the microRNA complement is often needed too. Just like a turnoff is most often needed for FMN (although sometimes aiming for the aptamer gate will do too.

The microRNA/s - here being more of FMN equivalent, since we still have the MS2, are still wanting a complementary stretch - just like the FMNs in the exclusion labs. This time for bind up to catch the microRNA, instead of turnoff and inactivation of the FMN aptamer. And this complementary stretch has gotten a turnoff sequence (identical or close to a part of the microRNA sequence itself).

And sometimes the microRNA and the MS2 can get made to share turnoff. (By word complementary gaming) Happens in the Sensor for HSA-mir 208a that we have had most winners in so far. Just instead of as in the Exclusion labs, where the MS2 demands a turnoff next to itself and the FMN is a good stretch away from the turnoff in sequence, in microRNA labs, the turnoff can be right between the MS2 and the microRNA complement.

MicroRNA turnon blueprint from winning design based on jandersonlee’s round 1 top scorer

MicroRNA turnon blueprintpng

MicroRNA turnoff blueprint in winning design by Salish moding Mat moding jandersonlee

This particular microRNA has some hairpin stem like features - like a stretch of A bases that can become loop. And here its two complements gets used as turnoff for each other to kick the microRNA off.

As something new, the static stem is kicked outside of the switching area, and likely aren’t needed at all. :)

The microRNA’s with two inputs do seem to get a somewhat T kind of shape, or seagull as I used to call it. :) And the turnoff labs have it, with just one input.

So I think the microRNA labs do have their own kind of blueprint. While part of it resembles partial moving MS2/FMN switches, a part of it is just microRNA specific fingerprints. :)

Single base microRNA catching dangles, no static stems in the switching area, can’t be locked in a circle except by hydrogen bonding...

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

Thoughts on the new lab results versus the MS2/FMN blueprint

The blueprint shines through strongest in both of the NG 2 labs. There even is the shade of a new blueprint. One that is probably equally forced both by aptamer orientation and by the designs having more bases than past similar labs.

I didn't expect Exclusion NG 1 and 3 to do as well as NG 2, but I had expected them to do better with my blueprint and excess bases kicked out. I also think they would have, had the aptamer been positioned just as in the NG 2 lab.

I think the twin G's in the FMN aptamer want to move opposite the stem they are closest to. I suspect this is actually better for our kind of switches.

However many of the high scoring designs also in the NG 1 and 3 labs keep recognizable shapes that resemble the MS2/FMN structure blueprint.

Aptamer direction matters

I have one more reason why the aptamer direction matters. One of the aptamer stretches gets used for turnoff in the Exclusion labs. The turnoff sequence is either a direct match to both the MS2 and the one side of the aptamer or a compromise between them.

The Riboswitch on a chip labs so far have shown preference for which FMN sequence they target. So reversing the FMN, means that the MS2 turnoff sequence will have to target a different FMN in space or change bits of its sequence to get to the FMN sequence that matched in blueprint space earlier. So orientation of the aptamer will affect the question on MS2 gate and no MS2 gate also. Since the creation of those are also connected to the sequence of the MS2 turnoff pyrimidines.

Small artificial aptamers

Many of our early switches actually have reversed FMN orientation to our later Riboswitch on a chip lab.

Which reminds me, we need some kind of way of talking of the direction of the aptamer. :)

Here are a few examples of designs of that type which would rather switch at their hairpin loop end than towards the middle of the design.

Also I realize that some of our old switches do have their FMN aptamers reversed compared to our latest Riboswitch on a chip. Even the NG 1 and NG 3 ones do have their aptamer reversed compared to the NG 2 labs.

I think this reversed orientation is what I seen the scientists use for several of their small artificial FMN switches and they typically tie the one end with a really small hairpin stem. (Image example). Search for the section in this post called Small switches:

Omei sent me this comment back:

Re the "inverted" variation, I find your example from the literature very interesting.  What especially caught my eye is that the closing base pair at the "top" end is UA, whereas Eterna players almost always use GC or CG.  I wonder if we've been missing an opportunity for improving the 3D shape of the FMN landing pocket when that end is bound in the NOFMN state.

I see that the orientation of every element matters. I think orientation of the aptamer do matter. If we get this puzzle piece answered on which side is most effective, we can make better switches. Even if it is just a 10% effect on fold change, it matters.

While I'm a strong proponent of one static end, I don't think having a few pair around the so called static end of the aptamer is a no go. I suspect it can be helpful to raise entropy a bit which again will help get things switching - if the switching area is otherwise small. What is important in our switches is that the two strands holding the aptamer is locked up with each other from one end. Then they can move a bit next to the aptamer if needed. However the reversing opens up for a new option with smaller switches where the whole aptamer moves.

I suspect the reverse aptamer as in the scientists artificial aptamers, will be far better for smaller switches where the whole aptamer is allowed to move and have much less space. It’s a way of raising entropy - make enough disorder - to provoke a switch. But still in a controlled manner - it being locked up inside a design.

What I find special about the example I sent you and several I have seen is that the "static end" of the aptamer is so small, that I could very well imagine that it is actually moving. However it being short, it greatly reduces the landscape for possible ways of forming, meaning that it can be a full moving aptamer, without the bigger penalty on it for having a hard time finding its stems again when it also have to fold. Plus small stems are easier breaking.

I found an earlier example of such switches with short ends, where I'm positive that there is a turnoff sequence hidden in the loop of this short "static" end of the aptamer.

Test of the blueprint

To bring my MS2/FMN blueprint idea fully to the test, I think the following will bring out the truth.

The NG 1 and 3 labs already has reversed aptamer in relation to the NG 2 lab. I think we should get far better lab results, if we get if the aptamer in NG 1 and NG 3 lab gets made as it is in the NG 2 lab.

Exclusion NG 3

Not winner

Exclusion NG 1

Ex NG1png

Not winner

Similar, since I think the Exclusion NG 2 lab has the right orientation on the aptamer already, we should get worse results if this aptamer gets reversed.

Exclusion NG 2

Exclusion NG2png

Winning design

The exact same test for the Same State NG labs as for the exclusion labs, should bring out the truth about aptamer orientation for them too.

Testing for size of switching area

A way to test if the static stem is as helpful as I think, is cutting out the static stem bases from the bases from the winning types from both Same State NG 2 and Exclusion NG 2. With one modification. I like the red frame that Omei put on top of my image. The static stem closing base pair will allow for the same structure.

Omei: Looking at your images, I had been thinking of the ideal comparison for cutting out static stems would be leave the closing base pair, e.g. the red rectangle.

Inline image 1

New NG 1 and NG 3 blueprint

A new blueprint have showed up in the Same state labs. I was talking about dangling stretches of switching bases, pyrimidine in particular earlier, because I thought I noticed a pattern in round 1 of the NG labs. Now this isn’t necessary a good blueprint. Except the bit with using both FMN sequences for a pair up at once, as that part is also in PWKR’s high fold change design and its mod sibling that took over the Same State NG 2 pool. That bit I think is worth looking out for in the future. (I did a separate analysis of the PWKR design type.)

I think this new blueprint gets allowed and favored due to all the extra bases in the NG labs compared to the Riboswitch on a chip labs, so that a stronger magnet section needs to get played. The SS NG 1 and NG 3 labs are basically going towards double magnet system, just like some of the past Same state 2 labs.

Same State NG 1 and 3 blueprintjpg

Notice this mostly pyrimidine stretch after the aptamer gate.

98%, by Mat

Most of the higher scoring designs carries a similar pattern. Usually the multiloop ring area isn’t so fond of having much other than A’s, so this frequency of pyrimidine is different. Especially because this trend turns up across labs.

96%, by Jieux

This is particularly related to the labs with the reversed aptamer, not the NG 2’s, while PWTR’s design has its own blueprint. (For attraction pattern I drew a mix between the folding engines, as I don’t think any of them got each state exactly right.)


Even the Exclusion NG 1 and Exclusion NG 3 has this pattern with a stretch of pyrimidines short after the aptamer gate. Which is unfortunate and why most of the winning designs break open and move away from become partial moving switches, but more half open switches, making it harder to make winners. As both Omei and I agree on any dangle left unchecked in single base area, will not be beneficial if its not doing something. Here is is trying to make a switch, but in an unhelpful way.

Else I had a period where I was a firm believer in the pyrimidine dangle as switch maker in the multiloop ring. Partly due to the round 1 NG labs.

One of the top scorers from NG 1

87%, by Mat

Here the pattern is forced by the only good positioning of the MS2. But still the result is a half open aptamer as the highscorers, which I think is less optimal.

Hyphema, 93%

I think that the Same State NG labs are far less hurt by aptamer direction compared to the Exclusion labs.

Perspective on FMN aptamer orientation

The reversed aptamer in NG 1 and NG 3 labs simply forces a different switch pattern. The FMN in reversed mode, wants a pyrimidine stretch very close to home to get anywhere. Where as we with the MS2 already have quite a lot of distance to the FMN aptamer, and as such want an FMN aptamer that is capable of doing a long distance pairing. I simply think that the FMN is magic in another way, from one end it loves doing short distance pairing, from the other end, longer distance is possible.

Big thx to Omei for questions and general discussion.

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

Two static stems in the switching area

Many of the Same State NG 2 winners in are related with a certain Xeonanis design, that originally stems from Nando’s bot ViennaUTC that scored 94% in Round 2 of Riboswitch on a chip.

Many of us modified it. Omei and I moved it along with us when we transitioned from the smaller Riboswitch on a chip designs to the bigger sized NG designs.

Back in NG round 1, I was speculating about the role of the static stem in the switching area. I tried to both prolong and shorten it in attempt to get data on this. As I shortened it, I realized that I left a lot of unpaired bases in the multiloop ring area, so I added an extra static stem before the MS2, to deal away with the extra bases. So the design had 2 small static stems in the switching area - instead of the longer one they replaced. (List of my designs, the shift between 1 and two static stems happens at mod 73-74. This new pattern with two static stems scored well in Round 1 and ended up hitting strong through in Round 2 among the winning designs.


Same State blueprint - now with more symmetry

Notice the equal distance between the two static stem in the switching area and the FMN aptamer.

I’m laughing of this dash of symmetry. In static designs symmetry in bigger amount causes trouble. Which again reminds me of the tandem glycine riboswitch with beautiful near symmetry. Both in secondary and 3D structure.

Many of the NG 2 winners carry this same symmetry. It is even the same in most of the Same State NG 1 and 3 high scorers.

96%, by Jieux

98%, By Mat

Hiding your excess bases

So the PWKR designs makes an extra aptamer loop that can’t pair with the first aptamer loop, since it is identical and already of rejecting base stretches. Where the Xeonanis pattern has two close to identical hairpin stems. Basically both things makes additional bases go away. But most usually other designs gets rid of the excess bases it by making one static stem.

Actually this Xeonanis design have far fewer mutable bases in the switching area (10) compared to the average NG 2 and Riboswitch on a chip winner. Usually there is around 12-14 mutable bases. Ok the range can go from 10-15 and even further.

While designs like this one with 20 mutable bases in the switching area also scores fine.


So while you can solve and get a 100% score without having 2 static stems in the Same State NG 2 labs, I think the bigger puzzle size makes it starting to get beneficial adding an extra static stem in the switching area.

But I find it interesting that there generally is a pattern for how many bases to hide away.

Size of NG 2 versus Riboswitch on a chip labs

The NG 2 lab has a bigger design size compared to the equivalent Riboswitch on a chip labs, Exclusion 2 and Exclusion 3.

For comparison the Exclusion 2 and 3 lab have 26 bases that are not MS2, in between their two aptamer sequences, whereas the Exclusion NG 2 lab has 30. (Same is the case for Same State NG)

Here is the highest scoring for the exclusion 2 NG lab. The MS2 is positioned next to the second FMN stretch. Notice that it is only 14 bases that are mutable in the switching area. Notice that more bases are hid away in the static stem (blue), compared to the equivalent solves in the Exclusion 2 lab.


Winner from Exclusion 2, that also have 14 mutable bases and a shorter static stem in the switching area.

Mean for Exclusion NG 2 winners with cluster counts over 20

(14, 12, 15,14)/4 =13.75 mutable bases in the switching area

The static stem in the switching area tend to be slightly longer in the NG 2 labs compared to the Exclusion 2 and Exclusion 3. Exclusion 2 typically have a 4 bp hairpin whereas Exclusion 3 typically has a 5 bp hairpin.

Function of the static stem in the switching area

So why was this extra stem allowed to happen in the Same State NG lab? I think I may have an explanation.

While the MS2 have to live with doing side way hugging in the Exclusion labs, it has always been easier to get the MS2 and the FMN to do a more direct hugging in a Same state labs (Same State 2 not included - since the MS2 was not held from both sides).

I think this extra added static stem in the switching area, besides getting rid of some of the extra bases, helps position the MS2 and the FMN in a favorable angle.

I think the MS2 prefers to get placed somewhat perpendicular to the FMN aptamer. Because it is then easier to get its C’s to hook up with the FMN G’s. Or as is the case for Xeonanis and the designs following the two magnet system blueprint, the G’s in the Aptamer gate.

Basically I think if or if not there are static stem/s, is directly related to the size of the switch bubble. If the switch bubble is microscopic - as in Sensor for HSA-mir-208a (MicroRNA lab) then a static stem is not needed inside of the switching area. However the more single bases there get in the switch bubble that is holding the switching elements available, the more critical it gets to make a static stem. So basically I think that the bigger the switch bubble is getting, the more needed it is to sprout extra static stems.

Now it starts make better sense why some natural occurring riboswitches have a wealth of what seems to be static stems in the switching area. :)

Here is an image of a natural occurring riboswitch - actually our old friend the FMN riboswitch. I’m guessing its stems will be stable as they look pretty solid. Except perhaps the orange colored one. (Looks symmetric too when split through the neck. :) )

Now with some more familiar bits like a sequence close to the FMN aptamer we know highlighted. It even looks like the aptamer is turning like in the NG 2 switches and our Riboswitch on a chip labs. And switching towards the middle of the design instead of towards the neck.

Switch bubble size

The difference in size between the switching bubbles for Same State 2 and Same State NG 2 and the effect of this small difference on the formation of static stems, made me realize something. That the size of the circle of bases that a switch holds, matters too.

Different sizes of the “ring” seems to affect whether or not there should be static stem/s in the switching area.

Which size bubble to choose for the switching elements, will depend on what is the aim with the switch.

I made a drawing to illustrate. Also this one is stylized. The static stems for the Same State NG 2 lab is right next to the MS2 hairpin.

Sprouting static stemsjpg

In the microRNA labs, the MS2 don’t need to be directly paired to the switching element. At least not in labs with one microRNA input.

A good size bubble - and it not being too big - has the benefits of raising chance for hitting on a good switch relatively fast.

MicroRNA blueprints

I realized that the single input microRNA labs do follow a similar style blueprint to that of the other switches, although it didn’t seem obvious in the first place. First they do not have a static neck, something which is crucial to a partial moving switch. Second, they have no static stem in the switching area, as a result they don’t holds a multiloop in both states. Thirdly, they are not partly moving at all.

Still they carry resemblance in their blueprint. Just it is a movable switch bubble with only the elements at static places, but moving in relation to each other. Still they form a bubble in which part of the switching takes place. Albeit it is a micro bubble - in the case of the turnon lab Sensor for HSA-mir-208a, except for the minority variant. In the turnoff variants a bigger bubble and multiloop and in the

Sensor for HSA-mir-208a

Majority variant originated from jandersonlee’s design

Mir 208a - Majority blueprintjpg

Type representative design

Minority variant originated from design by Brourd

Mir 208a - Minority blueprintjpg

Type representative design

Sensor V 3, Turnoff variant 2

Originated from jandersonlee’s design

Sensor turnoff 3 v2 blueprintjpg

Type representative design

Sensor V 3, Turnoff variant 1

Sensor turnoff 3 v1 blueprintjpg
Type representative design

MS2 gates in the microRNA labs

Now I know why the MS2 gate is so long in the early microRNA labs. Even the turn on lab (Sensor for HSA-mir-208a) had a long MS2 gate. Where normally the MS2 gate is usually only found in the turnoff (Exclusion) labs, if it is even there. Usually also the MS2 gate is rather short or max 5 base pairs, but in the microRNA labs the MS2 gate was huge.

I think the MS2 gate turns up because of the MS2’s general need to get hold from both sides for effective turnoff, when it needs to be gone. But this is not possible in the microRNA labs with single microRNA input, and no switch circle bubble forming in relation to the switching elements.

So what the RNA design really does there by making a long MS2 gate to form a micro switch bubble area between the MS2 and the MS2 gate. And then when switching, it drags in and slides out sequence as needed. And this secures that the MS2 is force held from both sides of its sequence.

This is why word change gaming is so needed. It’s the tackle that makes the switching glide.

I think we may need the word change gaming for all single microRNA input labs.

I have described the switch mechanism behind here.

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes
Ladybug for FMN Aptamer Orientation Memorization

I have been going on about aptamer orientation for a long time. FMN aptamers aptamers in particular and them wanting a specific orientation in relation to the switching area.

Omei decided to take a look at my claims for himself and made a visual overview from the latest FMN/MS2 Riboswitch Structure: the Paper (Round 101).

First round run: These labs have unlike the other labs not been run through several rounds and as such have not equal chance of reaching their highest potential score. I think given time Inverted Exclusion NG 1 and Inverted Same State NG 3 that have primary FMN orientation will beat their Inverted FMN siblings.

I liked it so much I wished to share it. Omei kindly allowed it.

Here is Omei’s document Comparative results of different FMN aptamer orientations with further images and comments.

Children’s rhyme for memorization of favorite FMN aptamer orientation

In my country, we have a children's word game of picking up ladybugs and ask them to fly up to God and ask for good weather. It rains a lot here... :) And then wait for them to take off - which is usually rather quick.

I however decided to use this word game for something else. Make it easier remembering which way a FMN aptamer really wishes to switch and by that if it is optimal for the design.

As far as I know Ladybugs don’t fly backward. :) So this should make it easier remembering that the red bases (red dots on the ladybug) like to sit at the back of the aptamer and the aptamer likes to switch in the opposite direction - forth.

Here are an image of the FMN aptamers with the red bases highlighted and the switching area indicated.

Image taken from earlier background post Different orientation of FMN

A lot of the time if only the worse option (at the right side) is provided, the design desides to switch in an open ended manner if there are bases enough at the other end - that here are labeled as locked. So designs in such a lab will try switch in their favored direction or even become full moving switches.
Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

Favorite FMN orientation may partially depend on puzzle size

I decided to take a look at the smaller labs in FMN/MS2 Riboswitch Structure lab, through the same lense Omei used on the bigger ones. See the above post and my intro on how to use Omei’s Data Analysis Table to do science stuff.  

There are all in all no strong trends. I just found it fascinating that what is preferred seem to be leveling much more out when the size of the switch bubble go down.

Only the Inverted Small Loop State NG 1 and 3 stuck out by slightly preferring having the inverted aptamer. Same pattern was only true for 1 of the Same state labs with inverted aptamers in the bigger sized labs.General trend for those bigger lab was to prefer primary orientation of the FMN aptamer.

Other trends for small loop labs

  • These smaller sized labs don’t seem capable of reaching same high fold change or scores as their bigger sized siblings.

  • Which may be why they turned to following trick: More full moving switches in general among designs with high fold change.


  • Designs with inverted aptamer are still more likely to be full moving than designs with primary aptamer orientation, just as in their bigger sibling labs.

I will look forward to seeing if our now bigger sized CRISPR FMN/MS2 puzzles will continue the trend from the bigger riboswitch on a chip puzzles so far and prefer even more having my favorite orientation of FMN. Primary and Non inverted.

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes

Switch Structure Overview

I have through time mentioned for individual lab rounds what is the preference for binding for ON and OFF switches. What I want now is to highlight apparent patterns across rounds and different types of switch puzzles.

I’m using a modified version of Omei’s Data Analysis table for demonstration.

Not only are we hacking RNA. While we are at it, we might as well hack science. :)

Designs with MS2 reporter works different from designs with microRNA reporter

What I in particular wish to highlight is that having microRNA inputs against microRNA reporters, spark the opposite pattern to having microRNA inputs against a MS2 reporter.

It totally reverses the pattern of if an input goes next to the reporter or needs to be distanced to the reporter, depending on if it's an ON or an OFF switch.

Inputs and their relation to the reporter mentioned as view from the last state in the lab puzzles.

Overview of Binding Preference for Different Switch Types


Also notice there is a slight change in pattern, just depending on if one uses one or more inputs.

Coaxial stacking that Omei noticed to be useful for our lab (plus underestimated by the folding engines) for also seems to turn up more frequently in labs that has microRNA inputs and microRNA reporters. Since with microRNA reporters, the inputs can get close to the reporter.

Whereas the turnoff sequence for MS2 typically takes up one side of the MS2 and leave fewer spaces for coaxial stacking there. Still inputs can have coaxial stems on either side, for an aid in binding so there are still options.  

Sum up

Strongest pattern when a switch hairpin reporter like MS2 (or K4) is used:

  • Reporter next to aptamer in OFF switches (Exclusion)

  • Reporter distanced to aptamer in ON switches (Same State)

Strongest pattern when a microRNA reporter is used:

  • Reporter distanced to microRNA input in OFF switches (Exclusion)

  • Reporter is next to microRNA input in ON switches (Same State)

Photo of Eli Fisker

Eli Fisker

  • 2289 Posts
  • 518 Reply Likes
Aptamers with inbuilt selfturnoff - a special group of aptamers?

In the beginning of this forum post Rhiju was asking about about if we could interchange other aptamers for the FMN that was the main aptamer we worked with earlier. I said yes, that I believe so, if certain contitions are fulfilled

Now I think I have a piece more to the puzzle. I suspect there to be a special kind of aptamer, with a different behaviour. 

In round 107 I saw a new aptamer fold type among some of the better scoring designs. Namely for the tryptophan aptamer. This puzzle was the one doing best in Exclusion Tryptophan B. I made a scripted tutorial explaing about this strange new aptamer with selfturnoff.

The aptamer seems to hold a sequence inside itself, that can be used for selfturnoff. I think there will be more aptamers that belongs to this group.

I think this behaviour is something we need to look out for as we can harvest it for good. 

Folding pattern for MS2 and aptamer without selfturnoff
  • A typical Same state (ON) design has its aptamer and its MS2 pair up with each other. 
  • A typical Exclusion (OFF) design will share a turnoff sequence that will match both inside the aptamer (or aptamergate) and the MS2. 

Aptamers with selfturnoff sequence will not use the typical bachelors dilemma style that are usual for many exclusion labs, where the aptamer and MS2 have identical stretches inside of them and share a common turnoff sequence.

Aptamers with mirror sequences

Plus I think this type of aptamers with selfturnoff may want something different to aptamers like FMN that can never turn itself off with the sequence it has, since FMN has almost mirroring sequences. 

These two FMN sequences don't have any complementarity with themselves.

What aptamers have in common

1) Magnet segments like GG and CC, also regularly more than one set

2) Repeat bases in general. 

This goes both for the aptamers with mirror bases as FMN and the tryptophan selfturnoff one.

The tryptophan aptamer actually do have some mirror nature to it if you watch the magnet segments. CGGCCGCC and GGACCGGG, if watched like this GGCCG. and GG.CCGG The magnet segments swap partners between off and on state.


Mirror Magnet Segments

Which brings me to something different. Mirror magnet segments

FMN, Tryptophan, Theophylline has them. 






Aptamer fusion

Now I wonder about something else. I wonder if I could stick this GGCCGG and GGCCGG sequence onto an FMN and have it work for selfturnoff? Aptamer fusion. :)

Not that the FMN isn't working pretty good on its own.

I bet we could make it work without the bachelors dilemma thing if we added the tryptophan "switch" sequence to the FMN. 

I also realized that we have a yet opened lab with two FMN labs. Albeit without the MS2, but still. I wanted to test the tryptophan selfturnoff mechanism together with a FMN. 

I ended up inverting the magnet segments above and stuck them together with a FMN in the still open lab. So it is GGCCGG instead of CCGGCC.

I made one of the red magnet segments in the FMN part of the one tryptophan aptamer sequence. Just as the magnet segments for one of the tryptophan sequenes is part of the tryptophan loop. 

So it doesn't look like a tryptophan aptamer, but the mechanism is similar. 

Exclusion design with tryptophan + FMN aptamer fusion. (FMN = locked bases, Tryptophan selfturnoff part = black rings)

Selfturnoff aptamer = bent aptamer?

There is one more thing I find interesting. The aptamers that appears to be able to be involved in selfturnoff are of the more bent (asymmetric) kind. Theophylline, Tryptophan, tetracycline are bent aptamers.

I found a potential inbuilt selfturnoff sequence even in tetracycline