The tiny URL for the table is http://tiny.cc/Eterna_R98_Fusion. I will redirect this to updated versions of the fusion table, if and when columns are added or corrections are made. If you want to go directly to this version, you can do that at https://www.google.com/fusiontables/data?docid=1FeYDTNyov47MUJPSVq4IzBit86j46_S1a_Ifx5Oo.
Big thx for getting all the sweet columns up in one go. Much appreciated!
Sending an applause to Meechl.
I have been looking a bit on the data. Here is a few thoughts on a series from the Same State NG 2 lab that had extremely high fold change compared to any earlier riboswitch on a chip labs. Only comparable to the best of microRNA labs.
First a few questions. I promise they will make sense later.
What does fmaxnoFMN_std and fmaxnoFMN_std mean?
From what I guess is that this has something to do with standard deviation and is not a data measurement directly, but more a measure of the accuracy of the measurement. But I really have no idea. And neither on how they are different.
Thoughts on a Same State NG 2 series
I might have been too quick on ruling out a pair up of the MS2 turnoff sequence with both FMN’s at once. :) I was complaining about EternaBot doing so earlier.
This design from Same State NG 2 has the highest fold change.
I have highlighted its pyrimidine and purine stretches. Notice just how many there are of them. Of cause part of them comes from the extra aptamer. However there is a relation between the pyrimidine and purine stretches. There always seem to be most purine stretches and fewest pyrimidine stretches. Even in the fuller moving switches, where the pyrimidine stretches run rampant. It even has a purine and pyrimidine stretch in the neck. I in general consider this more risky and better saved for the switching area. You want the switching area switching with itself, not with the rest and varying the frequency of the static stem in relation to the switching area, is a way of ensuring that. Especially since multiple stretches of pyrimidine and purines pose a risk for misfolding when in static designs. However I can’t complain about such fine results. :)
And same with a huge bunch of its siblings which are mod by Mat and JL mods. Ha, I see what PWKR did with his original design. He made an extra aptamer and placed it closer to the MS2. Way to go. :)
That's interesting. No static stem.
Now I’m wondering what goes on. Does the highlighted stretches form each their static stem and get the top aptamer and the MS2 really really close? (Like in the left image below?)
I’m guessing if anything is static it is that blue boxed stem.
Usually the static stems function in the Same state labs, seems to be to bring the C’s of the MS2 in close proximity of the FMN. Here a Salish mod of Brourd from Same state 2.
Here is a magnetic schematic of the PWRK design. State 1 flips in the opposite way than the magnet blueprints for the Same state puzzles.
Looks like there perhaps is an extra Same State blueprint. One that looks good for putting in the microRNA TB B instead of those two green plus stretches at the bottom of the circle. :) That’s if it can somehow live with a G turnoff sequence and if the data behind it is trustworthy. Which leads me to another thing that got me scratching my head.
I’m wondering about why this PWTR design and its siblings, had so high fold change compared to the other designs. More specifically I’m wondering if the extra aptamer actually bind an extra FMN molecule when in state 2?
I decided to check if any of the FMN column values were unusual in hope this would tell me.
I can see unusual numbers in the column under standard deviation for fmaxnoFMN_std for siblings of this design. Most of the designs with extremely high fold score - in range of the microRNA labs, are mods of this PWTR design.
Designs that have high fmaxnoFMN_std, also have high fmaxnoFMN_sem too. And not all that have high numbers are PWKR’s design. Now I wonder if these numbers in any way means that we can’t trust the data on these designs.
It seems someone has accidently left out the designer name in a column for easier translation of the data. I spend enough time designing these rna's, it takes a lot of time combing through the data to get my design's scores. As last round (97) One of the columns had designer in it and it was a wonderful google spreadsheet that didn't take up any of my precious time . Thanks and have a great Bird day!
Cooperative labs - round 2
I have been drawing blueprint drawings for the 4 highest scoring designs in the cooperative binding lab that we just received data on.
I already think a pattern starts to form, when also looking at round 1 where we didn’t have so much good data. Two of the best designs by Brourd in round 1, then had a static stem in the middle between the MS2 sequences and what I particularly noted - two static stem holding the MS2 well away from each other. (Cooperative lab data) I ditched the middle static and tried make a static neck next to the MS2 instead. As my first round results of placing a longer static stem in the middle had not worked well. So I aimed instead to make fully separate domains. However this didn’t turn out too well either.
Now however I do see the middle static stem between the MS2’s turning up again in a good deal of the higher scoring designs. Not that it is necessarily a strong static stem.
It's a strange lab. The base distribution frequency is quite different to usual. There is a huge % of U’s almost similar to that of A’s. (In the 30-40% range). Plus there is more C’s than G’s which is also rather rare. Normally there are more A’s than U’s and more G’s than C’s. Thx for the fine graphs, Salish!
It seems like the MS2’s make a direct pair up with each other for turnoff in state 1.
Just like the aptamer sequence needs to get anchored so it isn’t too far apart in space, the MS2’s are kind of anchored in space, but not always by a static stem, sometimes just by a big loop. This typically happens between the MS2 sequences, whereas the tail ends of the RNA sequence mostly tends to be loose at least in one state. I would still love to try tie them up to isolate the switch in a switch bubble. I tried in round 1 but not with good results.
MS2 not held from both sides
The top scoring designs seems to get solved in a half open matter. But often with one or two static stems. Now I have said that MS2’s wants to be hold from both sides when it comes to switches and in particular Exclusion labs, where the MS2 has to get turned off. However the top scoring designs in the cooperativity lab shows something different. That it isn’t necessarily needed for both MS2’s or even just one to be held from both sides. Now why this? This is a turnon lab. There is nothing a MS2 rather wants to do than turn on, and likely nothing it will rather pair with than its identical twin, as I bet it can make a pretty strong bind. It’s only turnoff sequence seems to be itself.
Early Cooperative binding blueprints
Pink = MS2
Blue = Static stem
Purple = Dangle (crossover)
Red = G magnet stretch
Green = C magnet stretch
The cooperative binding labs are turnon labs. And the magnet blueprints are in family with the Same state 2/NG 2 blueprints (Also turn on labs), despite the cooperative binding labs have no aptamers and not having a static neck. (Blueprints for Same State 2, see section Drawings of magnet blueprint segments) The magnet segments turn up at similar positions, especially when it comes to the double magnet system.
These 4 designs are using a double magnet system. (Two red and two green magnets stretches) As each MS2 naturally carries each one set.
This one is an open switch with only one static stem to the side. This static side stem tends to turn up at the same position in many of the switches. Just like in many of our other switches, it is a way to get the excess bases away. But I still suspect it playing a role somehow.
Most of the designs in the lab with the longer sequence length - Cooperative Binding - Multi MS2 - scored higher than those in the shorter lab - Cooperative Binding - Multi MS2 (shorter). This lab gave a fine chance to ditch both or one of the static stems, that there otherwise would be good space for.
This design I like because it has a static stem in between the MS2 elements. Which actually brings them close in space from one end. Just like were they FMN aptamer sequences, that were held close together by a static neck.
This one is full moving and while it is possible, I still think it is generally advantageous having a partial moving switch. Something that I think both MicroRNA and cooperativity labs will benefit from, despite they seem to be more open ended designs by their nature.
Notice how there is a cross over of sequences which separates the two MS2 sequences in between the MS2. It is tying a knot on itself, separating the two MS2 domains. Here helping the MS2’s turn on. A variant could also be imagined, a cross that made the first MS2 domain bigger than the second. Which might be interesting for reasons, related to a discussion which I shall mention in a post below.
Perspective on cooperative binding
I have higher faith in the one scoring 95% and the one scoring 91%. They both have special properties.
The one scoring 95% have a static stem between the two MS2’s, something I think can be a help both for a correct meet up when the MS2 needs to pair, but also for helping separation of things that should stay separate. Actually it has two static stems. Something I have been mentioning elsewhere also. (Static stems in microRNA)
I have special hopes for the 91% pattern. It reminds me of a conversation I have been having with Rhiju about the Glycine tandem aptamer riboswitch and kinkturns. Which I will bring below.
Discussion on Kink-turns
I’m here bringing a part of an earlier discussion with Rhiju on kink-turns and Glycine riboswitches, because it is related to the Cooperative lab and the new data we got. It all started here with Rhiju mentioning a different aptamer that is also a kink-turn.
I have found some good pages from the Lilley lab, that explains what a kink-turn is here.
(On glycine tandem riboswitches)
Date: Sun, 1 Nov 2015 13:52:58 +0000
Separately, I still don't understand why aptamer I is always longer than aptamer Ii. I suspect that it had to do with a conserved alternative state of these rnas but haven't yet teased this out. There is also our work on a kink turn leader in these rnas but again I'm not sure if that explains the structure or sequence patterns ...
On Nov 2, 2015, at 3:49 PM, Eli Fisker wrote:
I have been reading one of your papers (Automated RNA Structure Prediction Uncovers a Kink-Turn Linker in Double Glycine Riboswitches) on kink-turn in the linker region.
I found it very interesting. Giving you my raw stream of thoughts. Two main points.
1: I simply think you have discovered yet a method for keeping two domains separate. :)
First thought that popped up immediately, before I fully understood that this was not just an extra stem forming in the linker region, but a very special creature. (Good catch!)
I also thought about the ribosome and all the single stems I have seen in the linker regions and it suddenly all made very good sense, with all these extra stems that I had noted got formed between each domain.
It also reminded me of a video I recently saw on mRNA degradation and mRNA having an extra stem added forming at either end to prevent degradation. I simply think these stems between the domains have function, just like these mRNA code guard stems. But a different one. Protecting the borders of the different domains. And when thinking about it, this may also be happening not just at linker level, but in multiloops too. Stems working as protectors between sub domains (smaller domains attached to the multiloops)
Why I think this added spacer stem works: A strand mostly likes to pair with its nearest neighbor strand. (Crossover happens at a regular but far rarer frequency.) This is one of the main principle I design by for the microRNA riboswitches now. Making the content of a strand first aim to its one side in one state, and then to the other side in the next state, if I need to get it moving, turned on or off.
So simply adding in an additional stem in a linker region between two too identical domains, one ensures a good distance in sequence to nearest domain and thus much lowers the risk of the complementary parts in both the two neighboring domains from interfering. It is interesting that the glycine aptamers needs such a radical change. Now I wonder about other kind of tandem aptamers (I'm aware that glycine is probably the only cooperative one - at least as far as my knowledge goes) - do they have a far bigger degree of structural change and sequence change - if they don't carry a similar kink-turn? I could imagine so.
2: I think the kink-turn is an even more advanced way of keeping the domains apart, than just an extra stem for sequence spacing. The kink-turn simply twists the two glycine domains in different spatial directions thus making it even harder for them to meet in space and mispair - especially at the most critical points - the neck area - where they are also heavily altered in comparison to each other - both for structure and sequence.
I googled and found out that many kink-turns are already discovered in the ribosomes and from a quick judgement they seem to be more out in the branches and helping with protein interactions, and not being in a linker region.
Now I wonder how many other RNA dimers or other kind of RNA's that use kink-turns for their linkers.
I think there is something special about the glycine aptamer riboswitch that provokes the kink-turn at that position. It could as easily have used a stem for distancing, just like the ribosome. But it choose a twisted angle.
When I saw the 3D image in the paper, I couldn't help think that this thing is wicked pretty. It almost looks protein like in its symmetry and structure. It doesn't at all look like a messy RNA dimer. :)
By the way, it looks like the two multiloop in each of the riboswitches are close in space. Are they in any way touching or stacking or anything? If so could this in any way help add stability to the complex?
I'm guessing it is probably an angle thing. Just like with the switches in lab where I use a static stem to get the switching elements in a favorable angle in relation to each other. At least that is what I'm guessing that I'm doing and that the static stem has a function. Which leads me to wonder if we have any kink-turns in any of our switch labs.
Date: Tue, 3 Nov 2015 00:08:06 +0000
Thanks much for the detailed comments.
One thing that is different for this glycine riboswitch stem compared to the ribosome inter-linker stems (the ones you highlighted) is that the stem involves a rather long-range pairing (between the linker and a sequence-distance segment at the 5’ end) instead of a local hairpin — does that help you think better about domain separation?
On Nov 3, 2015, at 7:31 AM, Eli Fisker wrote:
Thx, this actually do help me. More thoughts back.
The A figure in the paper did make me see for a short moment that there was crossing over, but I forgot for the B figure where it doesn't stand out clear.
Ok, so the first aptamer domain is actually sealing itself off additionally, by the whole sequence forming "a closed ring", tying a knot on itself - a double knot. Making it far less likely that it will ever misfold with the second domain.
The first glycine aptamer domain already has sequence crossing over in space - the neck forming. With the kink-turn it is doing a double crossover at the beginning and ends of the domain. Ok, strictly speaking a triple cross - since the kink turn has two mini stems, besides its kink. It's unusual having this many cross overs in a row. Which is probably why it took longer time for it to be found.
In the Riboswitch on a chip labs I have long suggested sealing up the switching elements, by making a static stem (neck) with the tail ends of the RNA sequence. I make a cross over (neck) in those labs where ever possible to seal off the switching area - I also attempted to do this the impossible places too - Ex 1 and ex 4 - with expected bad results. :) And in Ex 5 and 6 with horrific results. I never before scored this many clean zeros. :)
I think this is why we haven't really got Exclusion 1 and 4 working well. (Ex 4 works better than Ex 1 because it has the MS2 turnoff sequence on the more optimal right side) Since they can't shield up and stabilize their switching area well enough, when only having the switching elements held from one end. I think RNA needs both "arms". Here an exclusion 1 design, with an MS2/aptamer turnoff lost in space. It really doesn't have good means of getting close to any of the two FMN's, although worse for FMN2 than FMN1, as it isn't sharing connected backbone.
The microRNA labs seems different. Which is why I go more by nearest neighbor strand there. They are probably having some kind of energy benefit from the long paring stretches between microRNA and their complements.
Do all kink-turns make such a far range contact? I could imagine those kinks in the ribosome for enhancing protein contact are not that far range, but mainly angle switchers, although I have no idea. I think the kink-turn in the glycine riboswitch has a double purpose.
I think the kink-turn has two functions. It is a triple insurance - on top of first structure variation and second sequence variance between the domains - plus it is a sculpting tool. For proteins it is beneficial to have their domains get close up in space - both to shield off hydrophobic regions in the protein. I wonder if there is an energy bonus or something beneficial from RNA being globular?
Glycine is neither hydrophobic nor hydrophilic, and as such don't seem to need shielding from outside environment. Not sure about the glycine riboswitch aptamer part of the equation.
Actually the kink-turn thing also means that I suddenly understand why this particular RNA looks so disturbingly much like a protein. I have just come to third chapter in my protein book. While I have read till now that proteins don't like to make big conformational changes inside their domains, proteins are mostly made of multiple units and those like to move around like with hinges between. Basically the kink-turn is a hinge for RNA domains. So I no longer find it as odd that I suddenly think RNA looks like protein. Just hilarious. :)
I read that the kink-turn can react to ion-content also. The glycine tandem riboswitch is a switch that needs switching between its domains too. :)
I think this might be related to my image with RNA needing both arms. It's not enough to seal off the domain with only one arm - linker sequence/stem. It somehow needs to get hold from both ends.
As I understand it, to make a kink in a protein, it mainly takes one amino acid, proline, to start a kink - more amino acids will be beside it but it is only one that makes the big time kink.
I thought RNA filled less in space than proteins, but I found a helix for DNA saying 20 Å in width, against "Ignoring the side chains, this helix is approximately 6 Angstroms in diameter". Not sure how much an alpha helix fills with side chains. Guess it depends on which.
I wonder if pseudoknots can be used for domain spacers too? I usually imagine them happening with fairly close by sequence. Not as far range as this kink-turn. Thinking about it, they may rather provide parallel angles between structure, instead of kinks.
Question: Could such a stretch inside the aptamer be a kink-turn too?
If so, I can find them in our riboswitches. And if so it might mean our static stem has a very specific purpose.
05-11-2015 Eli Fisker
Yellow boxed stems are different from usual necks. I think the neck is central for closing of a section of RNA. When first that neck stem has formed, this extra kind of stem more easily happens. The neck is setting the scene. Knotting the section off. Anything stem forming of strands close in space hereafter goes easier.
Which is probably also a reason why the neck needs to be longer in the first aptamer of the glycine riboswitch. It sets the scene for the kink-turn stems forming. I know I called it unusual the other day. Regretting that bit.
On Nov 13, 2015, at 4:50 AM, Eli Fisker wrote:
Thanks! Also for your pointers to the future with more aptamers playing together.
I need your help in relation to the L7Ae.
I have tried find it and I get nowhere.
What I need is its secondary structure and its sequence. Then I can compare it to the MS2 and the other aptamers for its interchangeability.
I think that many of those aptamers you mention can use exactly the same blueprint as the one I have proposed for the MS2/FMN, with very little variation. I have already been writing something about it in the forum a while back. But I think I can now explain it a lot better now, so it should save you guys some time understanding my pirate scientist language.
Wishing you luck.
Subject: Re: New reply: Blueprint of the MS2 and FMN Riboswitch
Date: Fri, 13 Nov 2015 23:53:54 +0000
L7ae binds kink-turns - so an ‘internal loop’ not a hairpin like MS2 — example sequence in here
In collaboration with the Greenleaf lab, we will be able to get the binding energies of L7ae to a variety of other ‘kink-turns’ over the next year or so, and should be easy to put into eterna. For now, though use that paper’s sequence. :)
On Nov 14, 2015, at 6:54 PM, Eli Fisker wrote:
Big thx for the paper. This saved me a lot of time and it was an interesting read.
As you have probably discovered by now, I have been having a lot of fun with answering your request. :)
Super short answer: Yes, I do think the riboswitch MS2/FMN blueprint, can be used also with elements interchanged.
Also now I more truly get what you meant by circular permutation.
No problem — thanks for the detailed posts! R
On Nov 21, 2015, at 7:07 PM, Eli Fisker
I'm getting happier and happier about the thought of kink turns in general. I think we can practically use them as domain spacers to separate up things that we don't want to have interfering with each other. Which should come in handy later.
[Particularly with thoughts on multiple inputs and switch elements. As soon as there are more than one switching element, I basically think that the need for static stems go up.]
So, initial analysis of cluster formation dependency in this lab follows known patterns for A, U, G, C percentages.
I believe the two six-figure numbers of clusters found are aberrations that can be ignored.
Interestingly, there appears to be a slight though significant dependence of the cooperativity on the cluster formation. In simple terms, the higher the cooperativity, the more clusters are created. While this holds true as a general trend, the bulk of designs scores up to subscore 22, up to which point the general trend holds true as well. As for all subscores higher than 30, it appears the actual values all fall short of the predicted trend. Then again, only five of us managed to get such high cooperativity points, so it is difficult to apply any real data analysis. On the other hand, 100% of these five values are repesented by data at >10 clusters, so they are valid designs.
As for the hillFmax,noFMN_MAD and then of course the K,D,noFMN values, the number of clusters decrease with their increasing values, as shown below.
And the cooperativity influence on cluster formation:
And the interdependence of kd,noFMN and hillFmax,noFMN_MAD on cluster formation
Sharing a discussion between me and Omei, because I think it can be helpful. It’s about cluster counts, rotated designs, plus turnoff sequence complementary, pyrimidine or purine style.
Date: Mon, 7 Dec 2015 21:57:17 -0800
Subject: Testing an NG 2 design against an inverted NG 1 or NG 3
I just thought I would let you know: I verified that at least one set of high scoring NG 2 designs can be converted into a comparable inverted NG 1 design. (Comparable here means the only differences between them are in how the static stems are sealed off.)
On Tue, Dec 8, 2015 at 6:14 AM, Eli Fisker wrote:
The ENG2/2-A2 design has in the low end of cluster counts - 13. But there is a sibling of it in the round 2 with good cluster counts.
It have the main blueprint shape and so does the NG 1 design - even though they have inverted aptamer. I know it is very well possible hitting the blueprint even with inverted aptamers, which is one more confirmation.
I just think we can get even better results with aptamer as in NG 2. Also there is one thing more. This design and its siblings follows the complementary solving style, which is less dependent on the aptamer sequence itself, as it hooks up with the aptamer gates instead. Which means these exact designs gets far less hurt by aptamer inversion. For exclusion designs not following complementary style, aptamer reversion will affect far more. (Those designs using a pyrimidine turnoff sequence for MS2/FMN turnoff) These designs did well in the Riboswitch on a chip lab for Exclusion 2 and 3. I got a winner in this style in NG 2. But not anything like it in NG 1 and NG 3 despite attempts.
Generally the static stem needs to allow for something like 4-5 bases in aptamer gate strand it is close to. Same state can do with 3 sometimes.
Notice that dangling tail in the NG 1 design you showed. The design has exactly the same amount of single bases in the multiloop area, by kicking the excess bases out as the NG 2 design. I generally got too many kicked out for the round 2 Exclusion labs. I went to far in the opposite direction, making some of my designs smaller than the original Riboswitch on a chip designs.
Date: Tue, 8 Dec 2015 11:12:22 -0800
Subject: Re: Testing an NG 2 design against an inverted NG 1 or NG 3
Thank you for your observations. I will certainly want to find other sets of designs where I can get an exact sequence match in the switching area. Do you happen to have a nomination among those you consider to adhere to your blueprint? Due to the small number of bases at one or the other end of NG 1 and NG 3 designs, the NG 2 design must have its interior static stem very close to the FMN aptamer.
One of your comments brought to mind something I might not have conveyed to you.
There seems to be little, if any, correlation of low cluster rounds between rounds. This graph comes from a group of designs that were synthesized in both rounds 95 and 96.
A design with a very high cluster does seem to be more likely to get a high count again in a subsequent round, so there does seem to be a reproducible phenomenon going on there. But low cluster counts seem to be more of random luck.
Also, I notice you often use "reversion" where I have used "inversion". The meaning of "reversion" is restoration of something to what it was at a previous time. That seems to have the wrong connotation. If "inversion" of a design doesn't have the quite the right connotation for you, how about "rotation". i.e. a rotation by 180 degrees?
The ones you found from round 1, are some of the ones that fit the blueprint. Its just as the NG 2 winners rotated. I like you found identical designs. Neither Exclusion NG 1 and NG 3 round 2 have real high scoring blueprints turning up, and I strongly suspect due to rotated aptamer. But round 1 showed they were possible, even with opposite rotated aptamer. Round 2 Exclusion NG 1 shows a few more of them than NG 3.
Here is one of mine from NG 1, round 2. It doesn't have impressive fold change. I think I got too many bases excluded from the multiloop ring and as single based tail stretch.
87%, FC 8.81, Fold change error: 1.06
And one of Brourd's:
Score 86%, FC 8.45, Fold change error 1.09
This one interestingly have a G turnoff sequence. I wonder if this a trend that could continue? [Earlier such kind of turnoff seemed to hurt] If so it may benefit Exclusion 3 blueprint type designs.
However most of the top scoring designs in Same state NG 1 and 3 show the blueprint. And again Same state labs tend to go more complementary solving style, so I think they are far less hurt by aptamer rotation, compared to exclusion labs. Which can be solved both by complementarity, but also a more magnet like turnoff sequence.
NG 1, Score 98%, FC 22.10, Fold change error 1.07
NG 3 winner, Score 95%, FC 17.68, Fold change error 1.08
Oh, 180 degree rotation I really like. Its very visual. The reason I say reversion is that our early aptamer aptamer labs are opposite to the riboswitch on a chip labs. We started out the opposite way. But I like 180 degree rotation. Rotation is more precise. I wish to have some way to specify which is which. I haven't found a good way yet. Perhaps in relation to what is normal orientation for nature. Usually FMN sits in the beginning of a messengerRNA. I think orientation matters a lot there. As it seems to also do for our microRNA inputs - which also lands at the beginning of a messengerRNA and favor pairing up with a specific end of theirs. But we are using the FMN in another setting, free of the messengerRNA setting and in connection with another aptamer, what is optimal for us may not be the same as in the natural setting.
Sum up of the discussion
I learned that cluster counts aren’t as much of an issue anymore since most designs have fine counts. And what really matters is how the aptamer turn in relation to the switch bubble. Not as much how it turn in the design sequence. All the NG labs so far have the aptamer turn same way in the sequence, but NG 1 and NG 3 designs have their aptamer rotated 180 degrees compared to the NG 2 designs.
Here is why we need to check from each direction of both aptamer placement in the puzzle and which way the aptamer turns in relation to the switching area.
The hidden hairpin loop in the loop ring
Earlier I noticed a pattern in a open ended Same State 1 design by Salish’s design, back in the Riboswitch lab on a chip rounds. It seemed as if the loop between the MS2 and the aptamer gate had a hidden hairpin loop, only showing up in one state.
I successfully transferred the pattern onto both Same State NG 1 and Same State NG 3 (Same State 1 equivalent). Same State 1 was hard to solve as a partial moving switch and it was harder to lock up the switching elements inside it. However the two later labs Same State NG 1 and 3 both had better options for becoming a partial moving switch.
The pattern seems to work best this way.
I later located the origin of the pattern back to Brourd’s design SS1 - 1011 from Same State 1, round 2.
Salish, you got me thinking.
I have long been wondering about why there turned more than one yellow hinge up in the same state designs. I ended up deciding that several of these often yellowish A filled stretches were helpful for forming either end loops, internal loops and multiloop ring. Plus sometimes switch between being either in a different state.
They seem to turn up at particular spots. The MS2 already have a hairpin loop hidden inside. But else these loop stretches tend to turn up at spots that needs to change identity with changing state. And be at the outer ends of those stems in the switching area in each state.
I decided to draw blueprint for the loop regions.
I have mostly ignored the static stem in the switching area in these ones. Also I haven't put much focus on the A’s in the aptamers, despite them being a kind of an internal loop.
I had drawn the twin G’s position into the aptamer. It got me wondering that I thought I had registered a difference in multiloop size for the 1 state structures. With bigger multiloop in Same State NG 1 and NG 3. To get an idea I drew up some main type structures from state 1 of some of Same State designs. (There are several different ones also).
This made me realize that the reversion of the aptamer caused a difference in structure types from state 1. I did a drawing to better demonstrate. Now we don’t have data on the inversion labs also the structures shown may not be the final best structures for that lab. But it goes to illustrate that the structures for labs that has their aptamer orientating the same way towards the switching area, are more likely to show close structural similarities.
I have an older drawing that I don’t think I got up, on Exclusion 2 and Exclusion 3 structure types for first state. Also they have structural similarities - while sharing same aptamer orientation. Both those structures that has a MS2 gate and those which doesn’t.
So I think reversion of the aptamer causes another structure in state 1, despite keeping the same structure in State 2.
Which isn't too strange when some solving types involves pairing up with bases in one or both of the FMN sequences. So of cause the position of the twin G's in the aptamers will influence the structure forming. The solving strategy that will easiest escape this effect is the complementary style as it doesn't always need to also pair up with anything in the FMN sequences, but can do with the aptamer gate.
I have earlier noticed there being a strong preference for closing base pair in the FMN aptamer. But I have been thinking about why the pyrimidine stretch or often some rather specific versions of it (CCUC) work rather well as MS2 and aptamer turnoff in the Exclusion NG 2 lab and in Exclusion 2.
The closing base pair of the aptamer simply allows for creating creates an extra G or C that can work in concert with the rest of the aptamer G magnet twins. I think the reason why it is working is that the pyrimidine turnoff stretch is directly complementary to a strong stretch in the MS2 and additionally complementary to an artificial prolonged strong stretch in the FMN aptamer.
Exclusion NG 2 - Pyrimidine Magnet Turnoff (Score 94%
Actually something similar is happening in the Same State NG 1 and NG 3 lab, despite the aptamer being rotated and there being no extra turnoff sequence as is costume in the Exclusion labs. It’s the same GAGG or GGAG sequence turning up, made out of the aptamer closing base pair. I'm aware it is also an artefact of aptamer gate complementarity to part of the MS2 hairpin stem too.
There is simply something about it that the MS2 loves, because it feels so familiar. :)
So it may simply pay making a bit of extra complementarity to the MS2, if there is not enough complementarity around with what can be squeezed into the aptamer gate. This is a story built over the same theme as the Hidden hairpin in a loop.
Pattern for closing base pair in the aptamer
This orientation of the unlocked closing aptamer base pair has been strong through many of the NG and Riboswitch on a chip labs. No matter how the aptamer was orientated and in which kind of lab.
They are really all the same FMN aptamer rotated differently in different labs.
Now the Exclusion NG 1 and NG 3 high scorers mostly have an AU closing base pairs and not GC pairs as shown above. However they were also mostly half open switches and there weren’t any winners.
I think what the Exclusion NG 1 and NG 3 labs really want is a GC closing base pair which is again in conflict with Exclusion labs wish to have the FMN as close to the MS2 as possible.
Really I think the end of the FMN aptamer with the open space for determining the closing base pair by far prefers having a somewhat static end or if it is to be involved in any switching, having a short stem. What is showing up in the Exclusion NG 1 and NG 3 high scorers are way too long aptamer gates. Something I think of as a hallmark of half open switches that finds it hard to get the switch going. Something that also characterized the Exclusion 1 and 4 labs.
Exclusion NG 2
In Riboswitch on a chip lab, the Exclusion 3 lab seemed easier solving than the Exclusion 2 lab.
But that wasn’t the case when we moved to the slightly bigger design Exclusion NG 2, that left us the choice where to place the MS2 sequence and we could choose to position it as in the Exclusion 2, the Exclusion 3 or something else.
The trend I noticed for round 1 in the NG labs with the Exclusion 2 blueprint beating the Exclusion 3 one, hits strong though and this time it is no coincidence. (Background post)
In the Exclusion NG 2 lab, which holds space so we can make both kind of lab solves, it is the Exclusion 2 positioning of the MS2, that comes out strong as the THE winner.
I highlighted the C’s position in the MS2 sequence. The small green boxes are the few exceptions that have the MS2 sequence next to the first bit of the aptamer sequence, equivalent to Exclusion 3 labs and these designs are not scoring in the winning department.
So this leads me to revise my blueprint collection. Earlier I considered the Exclusion 3 and its NG 2 equivalent best. But I see the Exclusion 2 style winning out. This spreads to Exclusion NG 1 (Exclusion 3 equivalent, just rotated design + rotated aptamer) and NG 3 labs (Exclusion 2 equivalent). Now I consider Exclusion NG 3 to hold most promise over NG 1.
I have noted with interest that I regularly can add different blueprints together. :) Even to get to a new one that I still miss. (I haven’t seen data yet for if that bit also works)
I have turned the glycine tandem aptamer upside down. I have highlighted what I judge to be static stems.
Now each of these glycine riboswitch domains which are close to identical domains only holds one switching elements. And as such don’t need to have their necks moving as our MS2/FMN riboswitches.
Our our MS2/FMN riboswitches however carries two switching elements inside them. I cut the FMN aptamer part out of the drawing of them. Notice that the glycine aptamer while going from 5’ to 3’ just like our puzzles, is mirror reversed to ours.
Seeing any structural likeness? :)
Glycine aptamer image from here:
Now we have an updated spreadsheet for the cooperativity labs.
After the update we got a new fine bunch of high scorers and even get to kept the former winners too. Simply can’t be happier.
Many of the new high scorers resembles this drawing I did of sub052, just with a shorter distance between the MS2.
Basically there are four main trends for the top scorers in the data till now. Small loop hinge between the MS2’s, static/moving stem between the MS2’s, MS2 domain variation and added extended complementarity between the MS2’s. I have drawn anew some design types for the sake of clarity.
MS2’s separated by loop hinge
In the lab with the longest design sequence (Cooperative binding - multi MS2), the yet simplest solve turned up. I find it fascinating that this works. Just a 2 base long hinge between the MS2. Even better, it's from one of the newer lab players. Congrats, christopherosep!
100%, Cooperativity score 40
MS2’s separated by static stem
After we got the new updated data, it still looks like having a static stem between the two MS2 could be an advantage. This is what one of JR’s winning design do in the lab with the longest design sequence. This was also looking good in two of the first round topscorers. (Link) Though the data from first round is shaky.
Score 95%, Cooperative score 34.70
MS2’s turnoff by hidden GC pair in an external loop
However something totally new happens too. All the winners by Brourd, have two very interesting things in common.
In the earlier spreadsheet update I noticed something in one of the top scorers.
Score 88.14 %, Cooperativity Score (CS) 28.14
Ha, I see what happened here. Yet another case of a hidden hairpin in a loop. Just also an external loop this time. Fitting since Brourd came up with that pattern in the first case.
Here it is even added from both ends, (State 1, left) both in the loop end but also in the exterior loop. (Thx to Omei for the exterior loop explanation)
However the new winners by Brourd show a clear trend. They prefer being only the extra bound GC pair by the top end of the design. Like this. The MS2’s have a hidden GC pair as an extension of the MS2’s pairing up and a weak movable “neck” of AU’s.
Score 100%, CS 40
MS2 domain variation
Hmm, this is really interesting. The one MS2 is not having the full MS2 sequence. :) One of the AU base pairs in the second MS2 is flipped.
This reminds me of something. I have long been postulating that it is easier to get two RNA sections to fold correctly if they are not fully identical. (Link) That is in structure and sequence.
Lately I have also said that double aptamer riboswitches needed domain variation (Link) In the tandem glycine riboswitch from nature, both glycine domains are not the same - they differ in both structure and sequence. Which I think is more needed as they are bigger than two small MS2 domains.
Where I tried to put the domain variation around the MS2 hairpin in the hope to not disturb it’s switching too much, Brourd went more radical. All the MS2’s in his winning designs, the second MS2 domain sequences differs from the locked MS2.
A switch bubble for the cooperative MS2 switch?
Here is one of Brourd’s winners that gives hope of locking up the cooperative switch inside a bubble. The other of his winners have a weak movable neck of AU’s. But with a GC closest to the pairing MS2’s.
Partial domain separation
I was particularly pleased that we got to keep JR’s Sub086 high scorer in the dataset after the update.
Notice the linker I have highlighted with purple in the image of the glycine tandem aptamer below. This is the kink-turn that Rhiju’s team discovered. I described more on it in the above post. This linker is binding together a stretch before and after the first aptamer, sealing the first aptamer off from the second. Something which I suspect help them avoid mispairing with each other.
This is also why I find this particular design by JR so interesting as it has such a knotting off of one of the MS2 section, although here it is the last one. While it still allows the domains to interfere and interact.
I would like to see a similar solve with the knotting up the first MS2 aptamer area instead. Like the drawing I showed earlier.
Now our cooperative MS2 switches seems to be quite different from the tandem glycine aptamer riboswitch, in that the MS2 seems to be very effective turning each other off, with no need for additional turnoff sequences and actual domain separation. Which do speaks for domains interfering with each other, instead of being separated as in the glycine riboswitch. The former did not look nearly as promising in the first round results.
Cooperative Binding Lab sum up - Round 2
The winning designs types show mostly basically the same blueprint - despite some structure variances. But with some bigger difference between the two labs.
The MS2’s likes to turn each other off. They can be separated by a short or a long loop hinge or by a static stem.
The shorter Cooperative binding lab seems to favor winners that have a short loop hinge between the MS2 hairpins. The winners have a distance of 5 loop/hinge bases between the MS2 hairpins.
The longer Cooperative binding labs seems to care less as it can have either a longer loop hinge, a real short one or a static stem between the MS2 hairpins.
Or put differently as overall for these lab, as long as there is either a static stem or a loop hinge between MS2’s, things seems to be good . :)
In the short sequenced lab, the designs gets extra aided in the switching by a GC pair that forms when the MS2’s gets brought together, along with a weak movable neck.
In the short sequenced lab, varying the unlocked sequences seems to beneficial, though winners without modifications are possible also.
I think the varying of the sequence in one of the MS2 domain works, because it weakens complementarity just a notch, while still being strongly complementary. So it both aids a not too strong docking plus departure. Adding strength for the pairing up of the MS2’s just outside may help to counter the weakening of the MS2 binding.
Something I think could be worth considering for the current open round is how much we can blend these blueprint patterns for the two labs together for the currently open labs. (Blending of blueprints) If the short lab will benefit from a static stem between the MS2’s and the longer lab will benefit from an extended MS2 pair up and MS2 domain variation.
I would also find it interesting if it would change the cooperative score noticable if it was the first MS2 that could be varied in sequence.
When I talk about complementarity system and complementarity style in FMN and MS2 designs, what I mean is designs that functions by having a stretch in the aptamer gate and/or FMN sequence, being complementary to the MS2. The principle is the same in other types of switches, despite them not having FMN aptamers.
By complementarity I simply mean that two strands of bases are matching up = complementary.
To get a switch switching, basically the switching elements or their surroundings needs to pair stretches of each other so the switch elements can either turn on or off.
Simplified complementarity shown. 4 identical strands that can match two and two, in two different ways.
A hallmark of switches is one or more strands that will be complementary to more than one place - which is how the switch is created. Nando described how to create complementarity in one of his blog posts New Year, New Switches.
Complementary style extended
This pattern hits heavy through the Same State NG 1 and Same State NG 3 topscorers.
By magnet system, I mean designs that mainly work by having 1 or more strong sets of magnets (stretches of C and G bases) that can be made to pair up and are causing the switch.
This is complementarity none the less, despite it being of a stronger and more magnetic kind and of less well mixed bases. Typically stretches of purine or pyrimidine. This is why I call this type of a pair up for magnet pairing instead of the generally weaker complementarity that happens when it is only sections of the MS2 that gets mirrored onto the aptamer gate.
Single and Double Magnet System
Magnets - what I call stretches of G’s and C’s when they are involved in the switching area. (Preferably in a partial moving switch). Double Magnet system is what I call the designs like the SSNG 2 types below, that besides having both G and C magnets in the MS2 has an extra set of magnets, in the aptamer gate.
The static stem, hinges and rotatability
Two of the major winner types in the Same State NG 2 lab, do not follow the blueprint like structural type that showed heavy up in the shorter Riboswitch on a chip lab. These designs have been bugging me. :)
These two types were made by Jandersonlee and PWKR, that later design I have earlier analysed. Both designs and their siblings are fuller moving switches - with only the neck area as a static stem. Both these mentioned design types have something distinct to them.
Purine Periodic Repeats - Excess magnet segments played
I found a new angle on periodic repeats which I saw heavily in the
first switch labs and ended up deciding was bad. It has popped up again,
but with an opposite sign and in winning designs.
The PWKR design has extended amount of magnet segments in play, by it having an extra FMN aptamer added in. This FMN pseudo magnet segments is brought real close to the MS2 in space.
jandersonlee mod of PWKR, Score 100%
This PWKR design has an overweight of purine stretches (G and A bases) compared to pyrimidine stretches (C and U bases). This is exactly the reverse pattern to many of our early switch labs, that tended to have an overweight of pyrimidine stretches compared to purine stretches instead. (Background post: Periodic repeats)
We have big trouble getting winners in a lot of our past classic and cloud lab switches.
I think a lot of them died since the structure and aptamer position in a good deal of the designs, forced us to to play pyrimidine magnet segments a lot more. The repeat patterns spreading from it, I think locked up the aptamer purine section too well in its off state, so I think it stayed more off than on.
I realized that the periodic repeats of pyrimidines, happened particularly in full moving switches, which we had a hard time getting good scores in. Partial moving switches seemed to have an easier time getting solved (Different types of switches) and later realized that by making a switch locked up by static stems, one could stop the bad periodic repeats from spreading too much on and it would then be easier to get it switching.
Whereas the PWKR designs instead play excess purine pattern and are super happy to get things moving. PWKR’s designs seems to be the most willing of all designs in the Same State NG 2 lab to switch. And this lab and its Riboswitch on a chip equivalent, Same State 2 have always been the easiest riboswitch lab hitting a lot of winners in.
What makes the PWKR designs different is that I think the repeated FMN aptamer, makes the design floppy, instead of the past switch labs that tended to make the design rigid by regularly having two pyrimidine stretches quieting the FMN aptamer purines by pairing with them - and probably doing the job far too well.
So basically this PWKR design type plays both a lot of excess purines and has almost full moving structure - and get away with it. :)
Which I find quite interesting. So in this cluster of similar designs, purine repeats seems to be far more happy to get things moving when they are plenty, compared to if it had been plenty of pyrimidine repeats. I wonder if we can repeat and make use of this somewhere else.
Also I would still like to know if this design can actually bind two FMN molecules, since it has two FMN aptamers. :)
Hyper complementarity - 3 gliding switch stems
jandersonlee’s design takes a whole other route. This design played multiple repeat strands and actually two moving stems on top of the the MS2, with heavy complementarity to the MS2. It has extended amount of complementarity. (Link to complementary style explanation)
jandersonlee’s NG - round 1, Score 100%
I have been wondering about why this switch by jandersonlee was scoring so well in the round 4 of the Riboswitch on a chip lab. It popped up in the NG labs and it and its siblings are growing strong among the winners. Just like like the PWKR design it had a rather big switching area - everything below the neck.
When taking a closer look, it turns out that this design is using complementarity, yet in a different way to usual. Instead of 1 static stem in the switching area, it seems to have two moving stems, one on either side of the MS2.
Mat mod of jandersonlee, score 100%
Vienna 2 says it is all moving where NUPACK and Vienna says the “green stem” is static. I checked a lot of it siblings and the mods and designs that breaks pattern in the right side stem so this stem seems to go static, generally scores a lot worse. So I trust Vienna 2 on this one. It looks like this stem isn’t static but actively involved in the switching.
This design works by having a strand next door neighbor to a stretch in the MS2 strand be complementary. What is very different about this design and its many siblings is that both the stretches meant to be complementary to the MS2 are not placed in the aptamer gate, neither are they direct complementary to any of the FMN stretches. (Aptamer gate - the stem forming at the end of the aptamer that is at same side as the MS2.) Which I think is worth taking note of, despite I have been annoyed at the design for fitting to the box of my expectations for riboswitch blueprints.
Another thing that makes this design stick a lot out from its fellow lab designs, is that it is usually only Exclusion labs (Turnoff labs) that has the MS2 turnoff sequence right next to the MS2. This is not a typical feature of Same state designs. This design has both a turn on and turn off sequence next to the MS2.
Actually this design makes perfectly sense. Repeat sequences and lots of them, are perfect candidates for avoiding a static structure and cause the structure to glide. Here just everything that is caught up in the switch bubble.
It seems important that the right stem after the MS2 is different to the MS2 stem stretch and the left stem. The designs where this is altered, resulted in lower score. I think this different orientation of the GC’s in the right stem, are responsible for the MS2 not getting tied too tight in state 1 and as such helping the switch moving.
Schematics of the design and its siblings
Skewed pairing in the switch bubble
Another of jandersonlee’s designs got me thinking also. This one follows blueprint type structure, but it had the magnet segments in the aptamer gate, reversed compared to the Xeonanis majority type among the winners. (Xeonanix - bigger design, JL’s design is the smaller)
Since I had realized that different orientation of the FMN aptamer was enough to spark changes in the 1 state structures, this made wonder this aptamer gate reversal would affect how the 1 state structure would look when compared to the majority type.
I noticed instead that it had the same 1 state structure type, as the Exclusion 3 lab had when the designs had a MS2 gate forming in front of the MS2.
Notice the more unevenly distribution of the strands that is to be paired up - in the switching bubble. They are paired up in a skewed way + jandersonlee’s design has only one set of magnets pair up in state 1, compared to what happens in the Xeonanis type design. Despite them having a similar amount of bases pair up.
I think the MS2 gate that turned up in some of the Exclusion designs will have a tendency to brake the switch and I think also this rather long switching stem in jandersonlee’s design may help explain why it is the minority type compared to the Xeonanis designs.
Just like it is possible in Exclusion 3 to make winners with a MS2 gate, it is also possible making winners with a real long switching stem as in jandersonlee’s design. But the more skewed the fold up gets, the harder it seems to be to make winners. (Something I have also earlier been pointing out.)
I think there will be a bigger yield of winners when we make the switching strands be somewhat perpendicular in their switching movement and have them fairly close together and evenly spaced. Plus cutting down on amount of unnecessary amount bases and elements and keep the moving parts limited.
Folding pathways - RNA origami continued
Basically the folds in the paper below signals how the blueprint type of our RNA switches likes to fold between two states. Somewhat perpendicular to each other.
A hinge versus a static stem
I have been wondering about the nature and number of the yellow Salish hinge. I have ended up thinking that it possible to have plural. Actually both the static stems and Salish hinges seems to turn up at specific spots.
The yellow hinges turn up at similar spots. So does the static stems. And they are actually overlapping. Or more precisely at the folds of the RNA origami fortune telling game. :)
Blueprint versus hinges and static stems
Basically the blueprint for a switch that will get working in a real short time and a small size are 4 strands that are complementary in sets of two and two and can shift partners. They also generally prefer being spaced out somewhat evenly and have hinges or static stems in between some of their elements.
Xeonanis type design shown at bottom
I see the blueprint even stronger. And it is similar for both exclusion and same state. They both attempt to fold into a similar structure via a similar mechanism. Just the positioning of the MS2 is different. As they are turn on and turnoff labs.
Xeonanis as a microswitch
The Xeonanis design which is one of the main winning types in the round 2 of the NG 2 lab, was perfectly perpendicular in the 2 state. It had 5 bases (7 bases if counting the base pair in the static stem also) on either side between the aptamer gate and the MS2 hairpin. (See the section Same State blueprint - now with more symmetry)
The newer structure of Xeonanis with two static stems in the switching area and just 10 bases that are changeable in its switching area (when keeping the static stems out of the calculation), shows that it is possible solving a same state switch puzzle very effectively in very little space and very few bases mutating.
Rotating the blueprint and function of the static stem
jandersonlee's design is fascinating in that it has similar structure in each of the states, but just kind of mirrored. There are similar sized stems forming to either side of the MS2 in each state. Similar is the PWKR design. Both keeping just the neck stable, but working well none the less.
Actually I wonder if they too can be rotated. I think I will try that out. :)
Hmm, I think there might be a problem there. They need a lot of space since they use up all of the SSNG 2 space, and there are not static stem in the switching area, that I can use for sealing up the design.
The static stem/s is really what is what is heavily abused for the blueprint rotation type. Creation of that static stem with the design tail ends is what makes it possible. :)
The latest Xeonanis design type with two static stems in the switching area, (Same State NG 2) compared to the Exclusion type winning designs that has only 1 static stem in the switching area, makes a difference when it comes to rotating the blueprint. Xeonanis is rotatable in all 3 rotation lab types, whereas even the Exclusion and Same state designs that follows blueprint are only rotatable in one additional lab.
Something clicked into place when I realized what jandersonlee's design was doing. It isn't rotatable and neither is PWKR's. Leaving the designs mainly following the blueprint or at least part of it over as main types.
Not that the other designs are wrong. They are possible, just they following another set of rules. Both jandersonlee’s with a very strong dash of complementary strands and PWKR’s designs with excess of the purine kind of magnet segments.
Size of the switching area
The limited and locked up switching area is what I find most interesting doing experiments in. Like the adding of GU's, at AU's places or reducing GU's to AU's. In already successful designs it will pretty quickly show if something is beneficial or not and over an array of designs.
I also know where I preferably want the switching area. The switching area should preferably be at the aptamer end that is closest to the MS2. The other end of the aptamer should preferably have a static stem. It has preferences for its GC closing base pair and a particular orientation, if the unlocked base pair is at the locked stem end of the aptamer. Just how the aptamer is rotated towards the switching area, determines a lot of how it is best solving.
What I find fascinating is that the labs and designs that follows blueprint structure, is that they have similarities even across turn on and turnoff labs. It is basically just 4 strands, switching into stems, two and two. These strands seems to like to be placed so they can do a head pair up on in 1 state and then do a 90 degrees change and then do a head on pair up.
Perspective - What's the point of an RNA switch blueprint?
I think keeping a design in a certain size and the switching area of a certain size, will spark mainly blueprint type structure. I think it is generally faster hitting on a working switch by placing the switching elements in a certain position in relation to each other, in a switch bubble that allows for strands to pair two and two and switch partners in an easy manner.
Basically the blueprint style of designs with the switching area locked up and only few and distinct areas moving is following the KISS method. Which nature has long time hit on multiple times when it comes to riboswitches. I see the same pattern in several naturally occurring riboswitches.
Or as my protein book so beautifully put it on evolution of protein oligomers:
“Evolution never ignores something that can provide a quick gain with little effort!” :)
(How proteins work)
I basically think the secret of making a switch work easily, is bringing the switching elements close in space and give them the right space/spacers in between them, be it loop hinges or static stems, to aid them move in an easy fashion.
Big end loops in stems for switch ability
I have earlier called out that it seemed to beneficial for hairpin stems that formed in the switching area, when the closing base pair of the hairpin seemed to prefer to be of a weaker kind - one that left the loop in the more positive range and unstable. Similar I pointed out that generally having a more positive loop - unboosted or boosted in a way that destabilized it - seemed to help the stem getting switching also. (see section HAIRPIN IN SWITCHING AREA)
I think I now see confirmation of that. I even think there may be a benefit to use slightly bigger end loops, as even unboosted tetra loops adds a stability on their own. If this really so, it should be beneficial, to move the MS2 turnoff sequence just a bit, to allow for a bigger end loop.
All the winning designs in Exclusion 2, Round 4 follows this pattern, with a bigger and more positive end loop. Whereas the tetra loop end loop turns up in some of the non winning high scorers. And a similar pattern follows for the few winners in the Exclusion NG 2, round 2. I also think I see it in the new XOR data.
Here the MS2 turnoff sequence is slided a bit compared to the MS2, often it has been right next to the MS2. I think it may pay off in some cases moving it a bit. Especially if one get the chance to make a hairpin loop closed by a GC pair at the end of this stem. Not all labs needs this. Exclusion 3 by far preferred having its MS2 turnoff sequence right after the MS2.