The most obvious case-in-point of this is getting array results for the Round 4 OpenTB designs. This has been an issue because Johan Andreasson, the post-doc who helped pioneer this technique and who personally conducted the Eterna array experiments, has gone on to a new job. He is still available for consultation, but can't spend the time needed to guide the experiment through its many steps. Fortunately, Feriel Melaine, the new post-doc who has successfully replicated the array results for the AK2.5 design, but with a bead-based experiment, has now agreed to take on the task of getting us array-based results for the Round 4 designs.
But in addition to that, we realized that we didn't really have adequate structure in place to do a rapid test cycle on the order of one every two weeks. The lab believes they can get their part of the work (receive the list of designs, order and receive the DNA templates needed as inputs to the experiments, run the experiments and return the data to players) in one week. But a two week cycle implies that players would then have only one week to look at the results, analyze them, disseminate that analysis to other players to create and submit new versions of their designs and then collectively decide on which designs should be submitted to start the next rapid feedback cycle. We have no precedent for having done that before. Now, we need to create a process to do that, and we'll be using the extended Light-up Sensors project as our testing ground.
Hoping to help organize the discussion a bit, I propose subdividing the discussion of the player part of the process into some individual steps. Starting with the next step falling into players' laps with the first 6 designs (currently in the lab):
- Players (typically the more experienced ones) analyze the results and try to distill what it "means".
- Analyses are disseminated in a form that is easily accessible by all lab players.
- Players submit new designs based on what they think they have learned from the analysis and subsequent discussion.
- Players select a new set of designs for the next rapid feedback cycle.
Zama’s Secret Sauce to spirals
As Zama said: I made that recipe looking only at the spinach- never liked spinach
Here is a lab discussion between Zama, I and later Omei too from a few days back.
Zama: @channel, here is the list for Trypto A SS Spinach. I didn't take the time to add links but the names are included and each is followed by a screenshot of the arcplot. I made a note if there were other similar mods. *The Secret Sauce*" for spotting a spiral (at least in this lab) is- *No multiloops in State 1, although a small hairpin near the static neck might be tolerated.* (edited)
Trypto A- SS- Spinach Spirals Google Doc - aka Zama’s book of spirals
Eli: Beautiful collection - big thx!
Zama: I'd love to hear everyone's opinion on which are best?
Eli: So far I like Dl2007's the best as they are fairly clean and very green.
There is one of Mee's looking real good too.
I'm basing this solely on color visuals.
The one of Mee's is the one with 0301 in the title
Zama: What about the last one- Astromon's?
Eli: It has some darker lines in the background on the top half of the arc plot
that may want to fold differently
But not bad.
One thing I notice is that many of these spirals have a little bit of circularized nature in them.
Zama: Can you define or give an example of what you mean?
Eli: Here is a more pure spiral by tone
I can't say for sure that the more pure version will do better than the slightly less spirally. Because the best FMN/MS2 aptamer with the best fold change so far for riboswitches, had an almost pure spirally pattern, but with a slight turn in the middle.
Zama: Can you give an example of what you call circular?
Eli: The ones that are like yin and yang shapes
if you draw a line through them, they are ring like as they can often be connected
while they are not round circles
more like salvador dali circles :)
Like this- with it setting on top of a spiral?
Eli: Example of circularized arc plot
And the design you show above is a spiral
let me see if I can find one of the mixtures
Zama: I saw tons of those- they are the ones with the multiloops. Were you saying you were seeing a lot of circulars in my collection?
Eli: No, Zama.
Just that a few of them have a slight hint of the circularized arc plot
Lots of designs in the labs aren't pure forms, but mixtures
Zama: Ah, that's where I got confused- and yes, some are on the border.
Eli: Well observed
And sometimes it also happens in the middle
Ok, here is one - actually one of those I like for their clean look.
I cheated a little and drew just the border
It has that yin and yang border feature
despite also being a spiral
Zama: Yes, those differ considerably to the kind of spiral we got in round 1- like Jieux's.
Notice that stem highlighted in state 1. This is the one giving the change.
Normally both moving part in both states of a pure spiral design will look very much the same.
They will be two or three switching stems in line. And all that happens is that they swap their partners for different ones between states.
Zama: Yea, that's one of those little hairpins in the static neck loop that seem to be tolerated.
The ones I were referring to- nice to know what they do
Example of the stems lining up in spiral designs:
Zama: Yep, no multiloops
Eli: Well put
I marked all the switching bases that pairs. Most of them become paired in both states
On the no multiloop observation, the spiral designs behave quite opposite to the circularized designs. The latter seems to make a virtue out of getting multiloops in.
Omei: I think the difference between a "pure spiral" and the "almost pure spiral" of the best FMN/MS2 designs is a very significant difference.
A pure spiral implies the absence of a multi-loop, and hence only a single hairpin loop. As I recall, one of the first things we measured in the earliest array labs were the number of hairpins in each state, and those with only one hairpin in each state, on the whole, were the worst group.
I think it is quite possible that pure spirals will turn out to have good fold changes, perhaps the best, but only if the experiments are done differently, i.e. allow more time to pass between changing conditions, to give the slower full spiral switches time to come to equilibrium.
This time factor (which the lab coats refer to as kinetics, as opposed to thermodynamics, is something that Eterna has tended to ignore, largely because the science of RNA kinetics really lags behind that of RNA thermodynamics.
Eli: Double aptamer design with best fold change 157
At first when I drew it up on paper, I thought it to be a true spiral, but it has a little detour from the spiral pattern in the middle.
I have been wondering if we can make a true spiral work in such a lab or if it would need a little longer sequence.
Omei: That would probably make it switch even more slowly, making it appear to be worse in the current experiments.
Eli: Yes, so a defect added somewhere strategic, while spoiling the pattern a little, may work the best.
Omei: I argued for Eterna to explicitly address kinetics questions. Rhiju was interested in talking about it, and came up with medical scenario where it would be important.
But there's too many other things going on right now, I think, to introduce that as new types of puzzles.
What I do think is possible is to start getting players talking about the significance of switching speed, so we can ask the experimenters to routinely publish the timing of the various steps for each round. If nothing else, that will give us a baseline for comparing results with future experiments that do explicitly treat time as an experimental variable.
Here are two inverted spirals from the Tryptophan B Same State (Spinach) lab
Counterclockwise spiral and clockwise spiral
Clockwise spiral (left spiral arc-plot)
Counterclockwise spiral (right spiral arc-plot)
This one is a near pure spiral, but has a tiny stem forming (aptamer selfturnoff).
The seed of the spiral orientation
There seems to be a pattern for the orientation of the spiral in relation to how the puzzle looks. The way the design bends around the turned off aptamer, hints of the orientation of the spiral.
- If the design bends clockwise, the spiral bends clockwise
- If the design bends counterclockwise, the spiral bends counterclockwise
Same State Example
In the Tryptophan designs it seems to be the tryptophan aptamer that is strongest and decides the orientation of the spiral. Same for designs with theophylline (at least those labs I have managed to make spirals in so far). Where as for the differently typed Exclusion FMN designs, it appears to be FMN that controls the spiral action.
Now this spiral pattern being closely connected with the design bend seems a emerging trend.
What I will be very interested in, can you guys make exceptions?
Can you make:
- designs that bends clockwise, but have the spiral bend counterclockwise?
- designs that bends counterclockwise but have the spiral bend clockwise?
Static stem deletion mystery solved
A while back I was wondering if the static stem that tended to turn up at specific spots in the switching area in lab designs were having a function. (Background forum post - See the section No static stem) Beyond the more obvious benefit of hiding away excess bases.
I got a chance to test this in round 101 where we got 8 sublabs with the FMN/MS2 riboswitches labs that had the aptamer placed such a way that there were no longer space for a static stem in the switching area, as the other labs allowed for.
I submitted sequences from earlier winners that did have a static stem, but just with with the sequence from the static stem part excluded. I expected the miniaturized designs to take a score hit. They all did, but one. So generally the static stem did seem to have a function. But that one exception have had me mystified since. I wrote about this design.
I can now explain why this particular design managed to escape the score hit dealt by the missing static stem.The static stem deletion caused a style change in the resulting design. Its arc plot is of the spiral kind. I think the spiral designs have a potential for reaching a far higher fold change than the circularized design style.
Spiral Arc Plot Needle in a circle Arc Plot Haystack
I was watching past lab winners in jandersonlee’s arc plot tool when I stumbled over this arc plot. It was a spiral. I looked up the design and realized that this was the design that had me baffeled from my #StaticStemDeletion experiment. It is from the one design from my static stem deletion experiments that had a better fold change than the original winning design that had the static stem.
I toggled the design from state to state and indeed this design a microspiral - a miniaturized spiral design. A shorter train of stems sliding. Watch how it slides. Just like the many recent spiral designs in the open lab.
Spiral designs with special switch graph signature
In addition to the spiral arc plot, the design had a curious switch graph.
I am wondering if the two curves not meeting at the right end of the graph image, means that the design has not gone through its full switch potential within the time limit.
I have found other designs with spiral arc plots that have the same switch graph signature that shares traits with my #StaticStemDeletion - JL SNG1 3.00 18 - variant 3 design.
Just to mention Zama's Tryp B SS (MGA) ZZ-1-18 and Jieux's Comedy 1 #all3 #staticstem from the lightning round 1 were also spirals, show the same switch graph behaviour.
Image borrowed from the newspost News on the Fast Track Experiments
It’s the same case for the switch graph of ViennaUTC's big spiral that gets the best fold change of all the small loops labs from Round 101.
It is possible to find spirals that don’t show this exact switch graph pattern. Like this switch graph for the spiral design by ViennaUTC from the partner lab. It doesn't have this switch graph curves as the other spirals. Best fold change of all the 8 Small Loop labs. Eg this one:
But I find it fascinating that the designs with spiral arc plots also seems so show early tendencies of trending toward getting distinct switch graph curves. The designs with spiral arc plot are different to the designs with circularized arc plots.
Many of the spiral designs (green boxes) had switch graphs where the blue and red line were apart and don't meet at the end (to the right). Just as the switch graphs I pointed out for earlier spiral lab designs.
I have added arc plot type to each design. Plus the #SB and #BS reveals the orientation of the spinach aptamer. B stands for big (longest side of the aptamer) and S stands for small (shortest side of the aptamer). That way I can use the same abbreviation for different aptamers. I use these hashtags on my lab designs, as for being able to group the designs after types.
So far there have been mainly counterclockwise spirals for the lightning round.
The patterns that seems to be there so far is that the Tryptophan B Same State (Spinach) with the best fold change and wide switch graph have in common, are that they are spiral designs + they use a selfturnoff binding for the tryptophan aptamer. Plus the two designs with the best fold change (27.7 and 19.5) + wide switch graph have in common, compared to the other designs, are that they have static stems at one end of both aptamers.
Arguably some of the Tryptophan A Same State (Spinach) designs employed the same strategies, but with less luck. The one with the best fold change (but also the lowest response) did both spiral and use tryptophan aptamer selfturnoff. (Jieux's Tidy Monster design)
Advice for round 5
If you are into spirals as am I, make sure you try out both clockwise and counterclockwise spirals. Plus try these out in combination with different aptamer orientation. That way we make better sure that we hit on what works best.
I have a proposal for voting strategy for Puzzles for lightning round 6, for those of you who are into spirals like me.
There are main 4 types of spirals:
1) Clockwise spiral (#CW) and big MGA aptamer sequence before the short MGA aptamer sequence (#BS)
2) Clockwise spiral (#CW) and short MGA aptamer sequence before the big MGA aptamer sequence (#SB)
3) Counterclockwise spiral (#CCW) and big MGA aptamer sequence before the short MGA aptamer sequence (#BS)
4) Counterclockwise spiral (#CCW) and short MGA aptamer sequence before the big MGA aptamer sequence (#SB)
While I try make a bunch of spirals in each of these 4 categories, I am not asking you to vote for my spirals. Pick whatever spirals you like best. What I'm asking is when you cast your vote for spiral designs, that you try pick one spiral in each of these 4 categories.
The reason I ask is that I hope is that we get a complete or near complete set of all 4 spiral types through the voting process. This may help start get an idea what orientation for spirals and MGA aptamers that may be best in these two labs. Which again may help us on what future lab rounds want.
zama: Many of our spirals resemble Fermat's Spirals. I'm curious if there is a way to gauge how close we can get to them?
eli: I think it has to with how you fill the loops out inbetween the sliding stems that are creating the spiral.
If you make an even number of loop bases on both sides, you will get a straight stem. I have a puzzle demonstrating.
Or rather I have 3 puzzles, that may get you a sense of it. https://eternagame.org/web/puzzle/9023368/
zama: @Thanks- and here I've been trying to avoid doing player puzzles lately- lol!
eli: In fibonacci spirals the spiral is growing, it isn't equally close all the way. To create a growing spiral you will need to add more single bases in loops on both sides in a growing pattern.
you can usually count on mine having something to do with lab.
Unless cynwulf has got me curious on something. :)
zama: The Fermat Spiral doesn't grow like the Fibonacci- why I thought ours looked more like a Fermat
eli: Ah, yes, I see I misunderstood
Ok, then you should be perfectly fine with even side loops on both sides of the sliding stems
as this puzzle
Actually the sides in the above image seems to get smaller (edited)
zama: Do you think the Fermat could be duplicated in design?
eli: I think it already has
That Pi design, as I recall it has the fermat patterns as what you showed
I wondering if the closer a design is to matching the Fermat equation the better the score??
That is a fine hypothesis
Now the spiral doesn't go the whole way through both of them. But there is a black strong line suggesting a bond will form and connect the two spirals. (with light green highlight)
Perhaps easier to see now:
Link to Pi's design: https://eternagame.org/game/browse/7559902/?filter1_arg1=7612876&filter1=Id&filter1_arg2=761...
zama: The design of yours that I posted above looks closer to me.
eli: Can you give me a link to it?
zama: It's from the google doc your working on. https://eternagame.org/game/browse/8787266/?filter1=Id&filter1_arg2=9037947&filter1_arg1=903...
eli: Mine only gets one full spiral through
Try play with the chain marker in the arc plot tool. First unmark, then choose chain. You can shift color by holding down shift
What I mean your fermat spiral image has two intertwined spirals both originating from the center of the sequence.
That is almost what Pi's design does
zama: Ok, it looks more like the Fermat image you posted.
eli: Yup. single fermat
I like this fermat illustration as the color contrast hightlights the inverted mirroring.
found it here: https://codea.io/talk/discussion/1770/fermats-spiral-another-experiment-with-large-images-meshes-and...
I see that the spiral arc plot you posted next to your fermat image is single spiral. But your fermat image is double spiral, just as my fermat image is single spiral. Both kind of arc plots are there. My arc plot represents the single fermat image and Pi's arc plot represent the two nested fermat spirals. (edited)
So really when we vote for spirals we should vote for 8 different types of designs. :)
4 single fermat spiral types + 4 double fermat spiral types
Recipe for double spirals
Ok, I'm not sure the double spirals are possible in same state labs. Pi's design is an exclusion lab. And there the aptamers have to counter each other. One be on in one state and another be on in another state.
Actually there are. They just look a bit different to Pi's exclusion double spiral. They instead get a characteristic snail shape.
Here is a good example from a previous post.
It is from Dl2007's design with fold ratio 1.13 (from Lightning round 4 results).
Watching Dl2007's design, I think the road to making double spirals is to focus on one of the aptamers. Then to pick two spots in this aptamer that can both pair with themselves in one state (1) and then have these two spots bind two places else in state 2, and have those two new partner spots in state 2 perhaps bind somewhere in state 1.
Kind of like this:
While we haven't seen high scores for double spiral designs yet, I would like to see them getting made and also voted for just as the same 4 types that I wish to see single spirals for.
@jandersonlee has added batch processing capabilities to his arcplot server, so it has now become realistic to process and evaluate large batches of designs. I've talked it over with some players, and we concur that it is time we start publishing some results to help collaboratively improve players' ability to evaluate designs.
The process will continue to evolve, but for this round (voting closing Wednesday PM), I am inviting all players (whether or not you are using arcplot) to post your recommendations for the two labs here. There are no strict rules for this, but I'll suggest the following guidelines:
- Limit your recommendations to no more than 10 per puzzle.
- Include your rationale for your choices, so other players can learn from your thinking. Depending on how you go about determining your recommendations, this might be a single sentence for each design or a general description of your selection process.
- Feel free to include a link or links to supporting material and/or to the designs themselves.
- Don't get hung up on precisely following these guidelines.
Theophylline A Same State (MGA) Arc Plots
I have an idea for multistate arc plot. (2+ states)
I will use a 3 state spiral puzzle I have made where all 3 states connectedly make a spiral - for illustration.
What I imagine is the arc plot as an opened book. Top page is the top arcplot for state 1. Bottom page is the bottom arc plot half for state 2.
Image by Katrine_S
Turn to next page and one could view an arc plot for state 2 and 3. It could also be 1 and 3. Or one could click an arrow to one of the sides for the page with arc plot for state 1 or 2, so one could swipe out one arc plot state for the next states arc plot.
Turn yet a page and one could view state 1 and 3 together.
Imaginary Arc plot tool with 3 states
Omei was wondering about why the Theophylline B Same State (Spinach) lab went bust, while the Theophylline A Same State (Spinach) lab went great.
I think I have an explanation.
A or B puzzle type?
A pattern that has been there from our recent lab with a reporter against an aptamer is the following:
When there are two partner labs, where the only difference is that the order of the reporter and the aptamer have position swapped for each other, then one of these partner labs tend to do a lot better than the other. There is an optimal order of aptamer and reporter in relation to each other.
In the R107 lab Single-input switches, revisited we had a lab round that looked at the MS2 reporter against a bunch of aptamers that were mostly new to us.
The aptamers were Tryptophan, Theophylline and Argenine. That lab too introduced the puzzle type naming A or B - in relation to if the MS2 sat outside of the aptamer. (A type) or if the MS2 sat between in between a split version of the aptamer.
Earlier we also had the R101 lab FMN/MS2 Riboswitch Structure. Those puzzle types weren't named A and B type. But what pattern has showed itself in relation to MS2 so far is that the B puzzle type generally has performed better than the A type. The pattern has been present from the earliest FMN/MS2 switches.
So Aptamer-MS2-Aptamer labs tends to go better than MS2-Aptamer-Aptamer.
Explanation to the image
The numbers 1 or 2 over the middle aptamer section signify if the aptamer is in one whole piece or if it is split.
The behaviour column where I have put MS2 equivalent to different aptamers, I'm basically stating which aptamer that takes the strongest aptamer spot, where MS2 would normally like to be if it was a part of a riboswitch.
Relative strength of aptamers in relation to each other
However now we are trying a bunch of relatively new aptamers to us, plus some new reporters. For these new aptamers I have mentioned that we still don't know which of them are strongest. As opposed to eg MS2 and FMN. In past labs MS2 has been pretty strong and rather stuborn, so that it took a good bunch of bases to make it turnoff.
I have expected the stronger aptamer/reporter to take the spot of MS2 - in the middle between the sequences of the other aptamer.
I will take a stab at future prediction for the fate of a couple of the labs that we don't have results on yet. I will base my guess on the result from their partner lab. The lab that has the same flourescent aptamer, but is of the opposite puzzle type. (A versus B type)
- Since Tryptophan B SS (MGA) went well, I guess that Tryptophan A SS (MGA) will go less well
- Since Theo B SS (MGA) went bad, I guess that Theo A SS (MGA) will go well
- Since FMN B Exclusion (MGA) went well, I guess that FMN A Exclusion (MGA) will go less well
If we are to do more lightning round labs before the full lab round, then I propose Theophylline A Same State (MGA) as candidate as I expect it to do well.
Request of extra lab puzzles
The Tetracycline Same State B (MGA) lab went bad. (Tetracycline-MGA-Tetracycline) But we never had its partner lab Tetracycline Same State B (MGA). (MGA-Tetracycline-MGA)
Provided that it is possible to make the partner tetracycline lab, I excect based on the result so far, that it should do better than the one we have done.
Basically I wish that we have a full set of partner labs for all the labs we have been working with so far.
I have noted that the two great designs, the one by Astromon and the one by JR both follow previous lab trends.
JR's good design is of the circularized arc plot kind - in this case looking like a clown mouth in natural view.
It is if the symmetric RNA origami style - making a switching movement typical of the majority of past winners. Plus both aptamers have static ends. Weird enough this circularized design has a similar switch graph as the spirals generally seem to have.
Astromon's good design is of the spiralized arc plot kind. It has an aptamerselfturnoff for the theophylline aptamer. Plus just as the 2 best designs in Tryptophan B Same state Spinach both of it's aptamers have one static end.
Theophylline A Same State (Spinach) summary
+ Astromon (9019376) #Spiral #ccw #bs #aptamerselfturnoff #two static aptamer ends
- Jieux (8895380) #spiral #ccw #SB
- Poll na gColm (9026438) #spiral #ccw #sb
- Dl2007 (8978861) #spiral #CW #BS
+ JR (8803108) #BS #symmetric Circularized arc plot #two static aptamer ends
#CCW - counterclockwise spiral
#CW - clockwise spiral
#BS - big side of aptamer first, small side later (for spinach or MGA)
#SB - small side of the aptamer first, big side later (for spinach or MGA)
+/- - weather the design did well or not
Recently I made a prediction for the few labs that we still hadn't done lightning rounds in from the Lighting Up Dark Cell's round.
Now we have gotten results back from two of these labs where I made guesses. Tryptophan A Same State (MGA) and Theophylline A Same State (MGA)
- Theophylline A MGA had good difference between the blue and red curves in several switch graphs.
- Tryptophan A MGA had only really minor difference between the blue and red curves in some of the switch graphs.
While the scoring scheme for the new switch graphs is still in the making and things as such are not set in stone yet, there are still things that can help us with deciphering the switch graphs.
Omei was answering switch graph reading questions.
Looking over the lightning round results, I noticed that on the whole, we have done considerably better with the Spinach aptamer than the MGA one. So I looked over some of the better MGA designs, like this one from the first round (Jieux's Comedy 1 #all3) to see if there was anything in common among those that did switch.
The first three I found all shared the characteristic of having an AU pair next to the larger of the aptamer loops
A quick check showed that this particular choice of pairs was not that common, and I wondered if there might be something in the scientific literature that could account for how an AU pair here could influence the binding. I found three experimental papers about the aptamer. It turns out that the bulk of the experimental data has been done on this hairpin sequence, with minor variations.
(Note that the base pair 11-22 corresponds to the highlighted pair above.)
The three dimensional structure of this aptamer binding to the malachite green molecule has been determined, and it is very unusual. The bound aptamer is stabilized by two base triples and one base quadruple. The first three panes of the diagram below mark the bases involved in these structures.
What is happening is that the RNA backbone at base 25 actual makes a U-turn, allowing base 26 to reinforce the pair U10-A23 and 27 to reinforce the pair C10-G23. Here's the 2D structure of the 3 reinforcing structures. (This figure comes from a paper that used a triloop hairpin instead of a quad loop , so the base numbers over 19 are shifted by one.)
Furthermore, the AUA triple has been shown to be very important for the stability of the binding pocket. Reducing the triple to a normal Watcon-Crick pair by mutating base A26 (A27) reduces the binding energy of the aptamer by 2.5 to 3 kcal, essentially eliminating its affinity.
Unfortunately, the question of most importance to us (should be always design with the AU pairing at this position, or are others just as good?) isn't specifically addressed in the scientific literature. There is a possibility that pairs other than the AU can form the same, or similar, triples. But the lightning rounds have given us results on 42 MGA designs that we can compare. Those results suggest that the AU pair is an especially good choice. Because there are differences among the rounds, there doesn't seem to be a unique way to make the comparison, but here is an obvious one.
Of the 42 MGA designs, only seven (17%) have used an AU pair in the same orientation as above. But of the six best fold ratios (> 2.6), four (67%) have used the AU pair.I intend to follow up on this over the next day or two, but I wanted to post this much now. The current lightning round is using MGA. and it seems that the choice of bases at this position should be a significant factor in both designing and voting.
The pair I am always referring to is the next-to-closing pair of the bigger aptamer loop -- bases 10 and 70 on the left and bases 59 and 27 on the right. (Regardless of the orientation of the aptamer, the stem base next to the closing G will always have a number one less than the G and the stem base next to the closing C will have a number one more than the C .) When I categorise a pair, I will put the base next to the G first. Thus, the next-to-closing pair in the design on the left is AU. On the right, it is UA.
For the purpose of calculating a summary statistic for a group of designs, one can't meaningfully average fold ratios. But the fold ratio can be converted to an energy value (called ddG) by taking its logarithm, and it is appropriate to average energy values.
So here's the average energy difference between states 1 and 2, grouped by next-to-closing pairs of the big end of the MGA aptamer. (I have omitted the UG and GU pairs because there was only one instance of each, and the values were unexceptional.)
This indicates that in the designs tested so far, the AU pair retains its advantage over others when considering all the designs, good or bad. As far as interpreting the significance of the ddG values, the difference between the ddG values for AU and UA (0.7) corresponds to slightly more than a doubling of the fold change.
- So how did the spirals do?
- Why do spirals work?
- Spirals and special switch graphs
- Spirals and aptamer self turnoffs
- Static stem deletion and spirals
- Spirals and static aptamer ends
- What is the benefit of spirals?
- Spirals and circularized winners and reuse of switching bases
- Spirals are the Schrodinger's cat of RNA
- Spirals and chain reaction
- Where have the spirals shown up?
- Spirals as top dog in labs that were hard to solve.
- Spirals and fold change
So how did the spirals do?
We have now had 8 lightning rounds getting feedback on lab designs. Michael made this real fine image (below) with overview of how well the individual lab designs switched, also in related to the expected max score.
3 designs exceeded the expected maximum ratio. (open dots over the orange lines)
2 of these 3 unexpected high jumpers were spiral designs.
Why do spirals work?
The first set of aptamers we worked with for riboswitches FMN and MS2 had the immediate advantage, that they had sequences that were a perfect match for a direct pair up between the aptamers, meaning that one could make same state switch where the aptamers turned each other off directly.
Similarly both FMN and MS2 had a shared sequence, which was perfect for using a shared sequence between them in exclusion puzzles, so the shared sequence could take turns with turning either the FMN or the MS2 off.
However many of our newer aptamers do not have the same shared sequence or directly complementary stretches.
The beauty of spirals is that they can work no matter if the aptamers are capable of matching directly or not.
What the spirals actually do is to allow an indirect pair up between aptamers that may not be able to pair up directly, like FMN and MS2.
Background on the match between FMN and MS2: The FMN piece inside the MS2 hairpin
Spirals and special switch graphs
Many of the designs with spiral arc plots, tend to have a special kind of switch graph, like the design has not yet finished switching and reach its maximum capacity for switching.
Here is the switch graph of Cynwulf’s maximum ratio overriding design:
Spirals and aptamer self turnoffs
Aptamer self turnoffs - which are aptamers turning themselves off by pairing with their own sequence, just binding in a different manner to when they bind their molecule.
Aptamer self turnoffs can themselves sometimes create spirals. This happens eg. with some tryptophan solves. Both the theophylline and tryptophan aptamer are heavily involved in aptamer self turnoffs and in causing spiralling.
Zama’s TRYP B SS (MGA) ZZ-1-18 from the Tryptophan B Same State (MGA) lab that was one of the 3 switches to exceed the expected maximum ratio, is both a spiral and has its tryptophan aptamer involved in self turnoff.
Cynwulf’s Set 2 Design 001 from the Tryptophan B Same State (Spinach) lab that was the other high jumping spiral design, also uses tryptophan aptamer self turnoff.
Static stem deletion and spirals
I was testing if a static stem really was necessary in FMN/MS2 designs, by deleting static stems in already successful designs and I expected these designs to fail.
One of these designs however did better than the original design with the static stem and this exact design had a spiral arc plot. Otherwise the static stems were beneficial to designs that holds circularized arc plots. Spiral designs can still benefit from static stems, however they prefer them somewhere different. Namely at their one aptamer end.
Static stem deletions to more regular circularized designs, may create spiral designs.
Background: Static stem deletion mystery solved
Spirals and static aptamer ends
While we have made spirals work without giving both aptamers one static end, I find it worth to highlight that the two spirals that exceeded the maximum ratio, both had two static aptamer ends.
What is the benefit of spirals?
Because size matters:
Spirals make allowance for bigger switches - and bigger fold change. This gets around the problem with too long stems that may arise that can kill a switch.
Because designs with circular arc plot tends to have stems. The longer stems get the less switchy they get. So there may be a limit to how big circularized switches can be. Switches with spirally arc plots can consists of several but smaller stems that are lining up as a train. They have the two sequences between their aptamers switch in parallel and a antiparallel movement. Whereas designs with circularized arc plots switch in a perpendicular manner.
Limits solve space:
While designs with spiral arc plots tends to be longer than designs with circular arc plots and as such uses more bases, there are still only are a limited number of ways one can make a spiral. Because the definition of a spiral is a bend in a specific direction. Plus the arc plot has to cross the line between states. So knowing that a spiral design can solve a design task, limits the space of possible solutions.
Image by Omei, background: The Maze of the arcplot
The spirals can be adjusted in length to fit the amount of bases one is given. A spiral can involve a direct complement to either 1 or both of the aptamers.
The switch may happen in a more gradual manner
As jandersonlee said: If you can 'roll' (or unzip/rezip) the base pair like this you don't have to break all the bases to change states. If the aptamer bases can switch first, they can perhaps free whatever blockage is preventing the reporter site from forming.
Spirals and circularized winners and reuse of switching bases
One thing that designs with cirularized arc plots have in common with the designs with spiral arc plots are that many of the same bases that are involved in base pairing in state 1 is also involved directly in base pairing in state 2 - just somewhere different.
Background: The maze of the arc-plot
Spirals are the Schrodinger's cat of RNA
Arcplot from Zama’s maximum ratio high jumper design. Notice how all the bases marked with blue are connected in both state 1 and 2 at least one time.
While the RNA can physically make what the spiral pairing up that is suggested with a spiral arc plot, the arc plot shows the potential for dual states.
The yet unfolded individual RNA bases are in a mixture of states - quantum state. With an option of being bound two different places. Although not at once.
Spirals and chain reaction
Some of these bases are even involved in chain interactions more than two places. Here are two chain interactions marked with different colors:
Spirals are basically a way of getting a maximum of bases involved in base pairing (of the switching kind) in both states.
Where have the spirals shown up?
Spirals showed up as as winners or near winners in labs before the lightning rounds
Small Loop Same State NG 3 (R101)
Inverted Small Loops Same State NG 3 (R101)
Exclusion NG 2 (R101)
Same State - Tryptophan B (R107)
Exclusion - Theophylline A (R107)
Same State - Arginine B (R107)
Exclusion - Theophylline B (R107)
Same State - Theophylline B (R107)
Same State - Theophylline B (107)
Exclusion - Theophylline B (R107)
Near winners with scores above 90%:
Inverted Small Loop Small State NG1 (R101)
Inverted Small Loop Exclusion NG1 (R101)
Small Loop Same State NG 1 (R101)
Inverted Same State NG 1 (R101)
Inverted Same State NG2 (R101)
Same State NG 2 (R101)
Spirals as top dog in labs that were hard to solve
Sole top scorer below 94%:
Exclusion - Tryptophan A (R107)
Sole top scorer below 90%:
Same State - Tryptophan A (R107
Exclusion - Arginine B (R107)
Same State - Arginine B (107)
Exclusion - Arginine B (R107)
Exclusion - Tryptophan A ((R107, Kissing Loops)
Exclusion - Theophylline B (R107, Kissing Loops)
A lot of the hard labs that did not have winners, had some of their topscorers show partial spiral behaviour.
Spirals and fold change
Not only do spirals turn up as top scorers in hard labs. In some of the labs where they have competition by the more usual designs that causes circularized arc plots, they bested the other designs by fold change, while not always achieving the top score. This is impressive, especially when taking into account that spiral designs up till the lightning rounds have been a lot rarer than circularized designs
Spiral designs coming out with top fold change for their lab (With fold change error limited at max 1.25)
Exclusion - Theophylline B design (R107)
Same State - Tryptophan A (R107)
Exclusion - Arginine B (R107)
Inverted Exclusion NG 2 (R101)
Inverted Same State NG 2 (R101)
Small Loop Same State NG 1 (R101)
Small Loop Same State NG 3 (R101)
Inverted Small Loop Same State NG 1 (R101)
Inverted Small Loop Same State NG 3 (R101)
Thx to Zama for catching some of my mistakes. I alone bear responsibility for the rest.
First a big thx to Zama for questions and discussion. Also thx to Omei for listening to my first rambling on double bachelor’s dilemma in spirals.
Zama has been asking me to explain more on what I say about racetracks, circularized arc plots and spiral arc plots. As she said: “...but there are circles in all arcplots so my head just wants spirals and racetracks and clowns to be a subset of circulars.”
She asked for visuals. She got a point, so here we go.
There really are a lot of circulars in a lot of arc plots, also the bad ones. But these arc plots tends to be a lot more irregular and swirly as Rhiju calls them.
They also tends to get really long. Which is okay if they are spirals, but not if they are not.
But Zama have a point of spirals having something really circular to them. They are moving in shrinking or growing circular movements.
Arc plots and no crossing
Circularized arc plots and spiral arc plots have something else in common.
Their road tracks don't cross over. Like in an infinity symbol
I have actually seen an arc plot for a design do something like this, although I forgot which.
Circular arc plots take the straight route
This is a circularized arc plot from a winning design:
Zama: The clown
Eli: Yup. Notice what the road do. It moves directly through all the aptamer parts.
Two times each
No extra detours
Zama: They lap back and forth
Design as shown in game, with the aptamer sequences highlighted.
Circularized designs that do less well, tend to take more detours
Spiral arc plots takes detours, although in a systematic manner.
Here is a winning spiral design (by Pi). Here the road track goes a lot different.
Design as shown in the game with the aptamers marked:
It goes from the theophylline aptamer, to a non aptamer area which again hooks up with the other aptamer (MS2). This is an indirect pairing.
I have been thinking about spirals kind of like being a double bachelor's dilemma in one puzzle.
When I mentioned the idea to Omei, he said: “I think I understand, and agree. To me, this kind of spiral seems closely related to the entangled pairs pattern.”
Tutorial puzzles on the bachelor’s dilemma:
Second part of spiral and second bachelor’s dilemma
It isn't always the two aptamers can share a common sequence like here.
But then if one can't hit the other aptamer in the third step, then one can aim for the aptamer gate instead. The aptamer gate can have its sequence changed to what is needed, unlike aptamers.
So one can always make a fit between two aptamers, by using sequence in between. Just like in the bachelor’s dilemma.
So basically the spirals either as single or double spirals works as bachelor’s dilemma. But there can be more than one inbetween sequence. But also fewer steps. The amount of strands between the two aptamers can vary. Depending on need. One can even make a spiral so short that one makes only a complementary sequence to one of the aptamers.
The difference between designs with spiral arc plots and circularized arc plots
So the main difference between designs with true circularized arc plots is that the circularised arc plots take the shortest route through all the aptamers (or really near them), whereas the spirals takes a longer and more indirect route. The later also through none aptamer or aptamer gate areas.
Spirals uses the indirect route to create complementarity between the aptamers, where as circularized designs do a direct pair up between the aptamers and or their aptamer gates.
The circularized designs depends more on short range interaction. Circularized designs do a more jumping move. Aptamers in spiral designs can be further apart.
Zama: So it's not that one is better than the other but what fits the design???
Eli: What fits the task. How long is the sequence. Is the aptamers fixed somewhere. Etc.
And if one want to make a switch that is likely to work with fewer tries, then one goes for a circularized design. It generally uses much fewer bases and as is easier to make.
However if one wants not just a winner, but a real high fold change, one should consider making spirals. Despite fewer of them are going to work and they are a bit harder to make. Because more bases, so more things can go wrong.
Zama: Which brings me back to the Round 5 comparison between Astro and JR's. One spiral and one clown.- The opposite of what you just said though.
Zama: This is JR's
Eli: Circularized with an extra arc
Zama: Yes, but in this case the results graphs are very similar and the circularized is the top design overall
Eli: Ok, I got what you are saying and you are right. The circularized designs is the best. For now.
I think the spirals have the potential to best circularized designs in the future.
Because I have seen it happen in a few of the past labs.
Zama: Me too!
Eli: So I had the mouth a bit too full and not saying that I was in prediction mode not exact fact mode. I sometimes forget.
Zama: I think the reason these two scored so well is because the spinach was about as far away from the aptamer as possible- on both. My hypothesis. Most of JR's reasons in his post didn't hold true in Astros.
Eli: This is a good observation. This also expand the space of the fold. It is generally helpful for same state designs.
JR's is a bit unusual in that it had its two aptames unevenly spaced. It has a bend. This is more typical for exclusion circularized designs
The design starts with a static stem, followed by the spinach aptamer, followed by the folding area, followed by the molecule aptamer , and finishes with another static stem. In this design the folding area slides or shifts the design to switch between states.
Arc Plot stats are as follows:
G11:C74 | S1: 0.57 kcal 38.63% | S2: 0.07 kcal 89.71% | ON 2.32x FMNgc
U18:A68 | S1: 2.92 kcal 0.79% | S2: 0.11 kcal 83.36% | ON 106.10x FMNgc
A28:U46 | S1: 3.06 kcal 0.62% | S2: 0.35 kcal 55.66% | ON 89.36x MS2
Astromon asked me:
@eli i have a question, that thing you predict which labs will do well (A vs B) will that spill over into the new labs?
eli: @Astromon, yes. I expect this to also hold for the new labs. One of the types will do worse than the other. Because aptamers have personalities. They are not equally happy about being hugged and held in between another aptamer sequence.
How I predicted the last set of labs
The way I predicted for the A and B type labs last time was that I already knew how one of them had done. If the A type has already worked well, then it was easy to say that that the B lab would fluke or the reverse. This has held for previous riboswitch labs.
No data on ATP yet this round
This round however we are having a new aptamer (ATP) that we have not worked with before. Also I do not have any of the partner labs to give a tip on which lab. However I have decided to give it a try anyway, but gone entirely differently about it.
Basically I think that:
- The ATP aptamer will do best when in between the MGA aptamer sequences
- The Spinach aptamer will do best between the ATP aptamer
NB, there is one of the Exclusion Spinach labs, which seem to be only solvable when the aptamers are entangled.