Help design the next round of lab puzzles!

  • 2
  • Article
  • Updated 3 years ago
Before running a final round of the OpenTB Challenge, we want to get the results for the first three rounds and have enough time for players to analyze them.  So the challenge for the upcoming round will be to design four small-molecule sensors.

The molecules we'll be designing sensors for are the MS2 coat protein and three amino acids -- tryptophan, theophylline and arginine.  The MS2 coat protein is an old friend, but this time we will trying to sense it (i.e. its an input) instead of using it as the reporter (i.e. an output).  The three amino acids are new to Eterna, and are perhaps the most exciting.  One of the most basic unknowns in the RNA World hypothesis is how proteins could have evolved from individual amino acids in the absence of the complex mRNA/tRNA/ribosome machinery that contemporary life depends on.  Demonstrating the existence of small RNAs that change folding when they sense specific amino acids could be a significant contribution to an explanation.

This much, the devs have decided on their own.  But there are some more options that players are probably more qualified than the devs to decide, and that is the positioning of the aptamers in the design, and the selection of the reporter RNA sequence.

There is still on-going discussion about the best choice for aptamer sequences, but here's more or less what they will be.

The MS2 protein and tryptophan aptamers will be hairpin loops.

MS2:


Tryptophan:


The other two will be interior loops, something like this:

Theophylline


In the case of arginine, I have only an illustrative example of an aptamer at this point:


Rhiju says it can be split into two halves, with free bases between them, but I don't have any more details yet.

The default for the reporter RNA, if players don't have any feedback, is the 10-base reporter we used in the first and third OpenTB lab.  This is Johan's preference.  However, he is open to using a different sequence of the same or shorter length.  The constraint is basically that he wants something that binds with approximately the same strength as the one we've used previously.  So it there are more C and G's, there may need to be fewer bases.  In any case, there can only be one reporter used for all the puzzles.

The choices we have wrt the aptamers is where to place them within the design.  Here, we can have up to two placements per aptamer.  But two placement versions for each aptamer will lead to a total of 16 different puzzles (assuming both a report-in and a reporter-out version for each aptamer), so if that seems like too many, we might choose to just have one positioning variant for some.

Since the question will undoubtedly come up as to whether we can make a puzzle where the aptamer can be "stamped" wherever a player chooses, the answer is not yet.  The current flash code requires that the aptamer (as opposed to the output complement) be at a fixed position. Nando is looking into whether he might be able to generalize that in future rounds.

So the discussion is now open.  Hopefully, the "lab" team members will get involved directly in answering questions about what is feasible from their point of view, and players will get their say wrt how to improve their ability to design the strongest switches. 
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes

Posted 3 years ago

  • 2
Photo of Brourd

Brourd

  • 452 Posts
  • 82 Reply Likes
Question: Why were small molecule sensors the experiment of choice to run during the interim period?
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
While the multi-oligo-input problem is being addressed by OpenTB,  dev team is completing a separate, first paper focusing on eterna's success on the small-molecule sensor problem. For that paper, we'd like to test if our understanding built up in 2015-2016 applied beyond FMN.
Photo of Brourd

Brourd

  • 452 Posts
  • 82 Reply Likes
Sounds good
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
OK, so now I'm changing hats and speaking solely as a player.  It seems to me that in the two cases where the aptamer is a hairpin loop, we only need a single placement, i.e. in the middle of the design. Even with the larger 25-base tryptophan hairpin, that leaves 30 bases on either side, and I can't think of anything that I would want to try that would require more than that.

If anyone does have something in mind that would take a longer contiguous stretch than that, please speak up.
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
super -- so now the question is: 
for the theophylline & the arginine aptamers, do we split or not?
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
(maybe we could do one puzzle where we split and one where we don't.) that wouldn't be too many puzzles...
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
The theophylline aptamer has fixed-8 bases on the 5' side and 11 fixed-bases on the 3' side, for 21 bases overall. If the working length for designs is still 85 bases, that leaves 64 bases to work with. If we want just one case, I might suggest splitting this fairly evenly as 21 bases for the tails and 22 bases for the internal free structure. If you end up insisting on a 3-pair stem on either end, that still leaves 19 base tails and 18 bases of free-form internal structure. It may be a bit tight but should be workable.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Here is a bit of input


Whatever reporter sequence is chosen for these new labs, its complement should preferably have at least 3-5 bases compmementary to all aptamers.

Also the reporter complement should preferably have something like 3-5 bases that are identical to base stretches in as many aptamers as possible. 


Fit between aptamer elements that have never met in real life


I recall when Omei found an example of the MS2 and the FMN both having a sequence fragment being an identical to each other. See the section The FMN piece inside the MS2 hairpin

The MS2 and the FMN together were magic together because both the FMN and the MS2 shared identical sequences as Omei demonstrated but also in that they had sequence stretch that was directly complementary for direct turnoff. This way one gets two ways for a shut off. A direct one where both MS2 and FMN will be off (or on) and an indirect one, where they can take turns being on and off. Not sure what the lab needs in that respect. (For more background, see the post I mention below)


1) Identical sequence strecth in aptamers and reporter 

Basically for the reporter and the aptamer to fit well for a turn off for both with a middle man sequence (bachelors dilemma), the reporter needs something like 3-5 bases identical base stretch to as many of the aptamers as possible.

Then the bachelors dilemma sequence can take turns with turning off the aptamer and the reporter complementary stretch.

Since all the aptamers have a good bunch of double C and G magnets, one such needs be among.

Tryptophan and MS2 have the following two sequences in common GAGGA (which is very FMN aptamer sequence like) . Theophylline has AGG and can be made to have an end A.

Tryptophan, MS2 and Theophylline has CCA in common.

Basically the best reporter complement will be the one fragment that most aptamers have in common.


2) Direct complementaity possible between aptamer and reporter complement

However for another type of turnoff to be achieved, a short stretch of bases that are directly complementary between aptamer and reporter complement will be very helpful.
Direct complementarity for the aptamer between reporter complementary sequence and aptamer sequence will ensure they can switch each other off directly.

The argenine and theophylline aptamer would love a reporter sequence that is complementary to GGAAC as it then would be complementary with both the reporter and the aptamer. Meaning the actual reporter would end with CCUUG. I did a demonstration in the post I mention below.

Another good fit will be GGUG as this will fit for turnoff of MS2, tryptophan and argenine. Possible also theophylline. While not a straight match for most of these aptamers, magnet segments have shown to work well even when having a mismatch inbetween strong bases.


Background article

This earlier post can practically serve as background for picking sequences that will help turnoff aptamers and reporter complements.



Flavor of the reporter

Also it will be best if the bases chosen are of the CU type or GA type. The more reporter is kept in GA (like FMN) and/or CU rich regions, (MS2 does both) the more switch like I think it will be. Also the reporter shouldn't be too hairpin stem like. (fond of folding with itself) Rather microRNA like - with base bias either towards C or G.
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
can you propose some specific 8-nt, 9-nt, and 10-nt sequences?
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
I have been drawing up the aptamers and highlighted their magnet segments to better make a decision on where to cut a sequence. 

What all aptamers have in common are at least one set of both G and C magnets. (Double/tripple C's, G's.) And even often more magnet segments.

For all these designs it is true that their G and C magnets comes rather close. Not always in same order. The Theophylline and Arginine aptamers have the magnet segments at the position where I would prefer to catch them, in reverse order to MS2 and Typtophan.

Since MS2 only has one set of double magnets it will decide the order. However both Arginine and Theophylline contain single G's that can compenstate for not having a double G.



I have marked the position of where the complementary or identical sequence will cover in the aptamers.

Here I have a new sequence both as identical and complementary that will be interesting for a strong turnoff. It won't exactly be identical in position and bases for all aptamers, but there should be enough strong bases for them to make connection.


1) Reporter with identical sequence to stretch in aptamers

Identical sequence that I would wish to see included in the reporter

GGXXCC

Or something like GGAACC.

The middle bases are less important, just they should preferably cover most possible of the aptamers. There won't be full overlap for the middle bases, and the distance between the G and C magnets in the aptamers won't be equally long. Yet the magnetsegments should ensure a pair up.

An example of a longer sequence including this could be:

GGAACCAG

GA rich

Carry close to identical sequence to aptamers. Can be continued in both ends with AGGA that is complementary to CCAC.


2) Complementary reporter

Reporter sequence that would be directly complemenartary to parts in the aptamer sequence.
CCXXGG
Or CCUUGG

An example of a reporter sequence could be this:

UCCUUGGA

Carries complements to the longer closely shared aptamer sequences, plus UCCCU is complementary to AGGA which partly show up in 3 of the aptamers.

CU rich

While both these early reporter ideas carries both a G and C magnet, they are so close that they are not in immediate danger of pairing with each other and forming a hairpin.
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
A 3-palindrome GA rich reporter with CCXXGG could be:

UCCUAGGAAG

contains UCCU, AGGA, GAAG.

A 3-palindrome GA rich reporter with GGXXCC could be:

GAAGGAUCCU

contains GAAG, AGGA, UCCU.

A 3-palindrome CU rich reporter with GGXXCC could be:

AGGAUCCUUC


A 3-palindrome CU rich reporter with CCXXGG could be:

CUUCCUAGGA


A 3-palindrome CU rich reporter (no A!) with CCXXGG could be:

UCCUUGGUUG

contains UCCU, UGGU, GUUG.

A 3-palindrome CU rich reporter (no A!) with GGXXCC could be:

UGGUUCCUUC

contains UGGU, UCCU, CUUC.

and so on.
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
Since no one else has spoken up, I will make the case (once again :-)) for a longer reporter.

The 10 base reporter we have used gives the designer almost no flexibility in the reporter complement sequence.  In contrast, the 14 base reporter that we used in Round 2 of the Open TB challenge gave lots of different options for introducing a bulge or mismatches, and players made use of that.  Not counting mods, the top 5 designs in Round 2 each used a different reporter complement:
(I don't have the data to substantiate it, but my gut feel is that the longer reporter made a bigger contribution to the success of Round 2 than the switch in oligs did.)

As to the perceived drawbacks to using a longer reporter, in the dev channel Joanne wrote
I still prefer the 10-nt reporter if players are OK with it. We need to be able to measure the true KDs (nM, not low pM) to even have the option of determining  the degree of thermodynamic optimality.
So I looked at the difference in the percentage of designs in Rounds 1 and 2 that had KD_ON values less than or equal to 1 nM.  In Round 1, it was 5.2% of the designs, while in Round 2, it increased to 11.2%.  (And indeed, all the best global fold changes fell in this category.)   

However, there has never been any incentive to avoid very low KD values.  I know Eli has talked about how getting KDON as low as possible is the best way to increase the fold change.  If low fold changes really are causing experimental problems, I think the best way to address that is to incentivize players by including that criterion in the Eterna score and let players know.  The baseline subscore already "docks" a design if its KD_ON value is too high; the obvious thing is to also dock it if it is too low.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Omei, you are bringing up something very important.

I would want Johan to get the data quality he needs to validate our designs correctly.

So first obvious question. Which KDON value will be too low?

As Omei points out, if we know what is too low and unwanted, we will try avoid.

Would it be possible with a compromise between what Omei suggest with a longer reporter and what Johan wants, by us not using more of the reporter than like 10 bases?

That would get us the best from both worlds, better data quality for Johan and more reporter pairing options for us.

I mean this way we get to stuff in more magnet segments, potential good identical stretches and complementary stretches between reporter and aptamer into the reporter and be able to call what part of it we need need for the specific lab.

OFF and ON switches doesn't always need the same. For the riboswitch on a chip labs, the ON (same state) switches trended toward a direct pair up between FMN (Or the stem region closing it) and MS2. Meaning both MS2 and FMN would be either on at the same time or both off at the same time.

Where as the OFF (Exclusion) switches trended towards using a middleman sequence, for taking turns turning either the FMN or the MS2 off.  
(Edited)
Photo of Andrew Kaechele

Andrew Kaechele

  • 85 Posts
  • 27 Reply Likes
"The 10 base reporter we have used gives the designer almost no flexibility in the reporter complement sequence." I agree with Omei here. 
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
In the past, we have talked about the desirability of a having a "universal" reporter sequence.  It's possible we have enough experience to home in on that, and perhaps we can find a way to actually test it in this round while still addressing Johan's concerns about KDON measurements that are too low for him to get precise readings.

The sequence I will propose is AUGCCUAGGUCA:


The characteristics it has that seem valuable for universal reporter are
  • It has the two magnet sequences that Eli strongly favors.
  • It contains all 8 possible F/Y triplets.
  • It doesn't form a hairpin on its own.
  • It has a few more bases that strictly necessary, to give switch designs some flexibility in selecting a good kernel attractor.
I'm not going to make any claims that this is the optimal choice; I doubt whether it is.  But perhaps considering what would constitute a universal reporter could provide some unifying ideas for deciding what we should use for this specific round.
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
@Omei said: "It contains all 8 possible F/Y triplets."

Do you mean all R/Y triplets, as in:

GUA => RYR
UAC => YRY
ACC => RYY(1)
CCU => YYY
CUA => YYR(1)
UAG => YRR
AGG => RRR
GGU => RRY
GUC => RYY(2)
UCA => YYR(2)

The last two bases yield duplicate in this regard and could be dropped to give a 10-base sequence which the length that Johan would seemingly prefer: GUACCUAGGU

However while your sequence has 3 each of ACGU, this would only have 2CA and 3GU. Swapping bases 1&3 and 10&11 before truncating would yield:

AUGCCUAGGC

with 2AU and 3CG.

However this does fold on itself as ..((....))

A 10-base sequence with 2AU and 3CG that doesn't (strongly) self-fold might be: 

AUGUCCGGAC  (RYR,YRY,RYY,YYY,YYR,YRR,RRR,RRY)

or perhaps the more symmetrical:

UGUCCGGACA (YRY,RYY,YYY,YYR,YRR,RRR,RRY,RYR)
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
Some 10-base sequences (the length Johan would seemingly prefer) that have all eight R/Y (puRine/pYrimidine) triplets and yet do not strongly self-fold include:

UGUCCGGACA


Or with fewer strong bases (As Eli mentioned elsewhere):

UGUCUAGACA


UGUUCGAACA


UAUCCGGAUA


These do not include Eli's 4-base palidromes.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Palindromic aptamer core  - Switches with mirror sequences


I have been doing some more thinking about the nature of what a good reporter candidate should contain and what a good aptamer turnoff sequence is.

I have been drawing some draft images of some of the possible scenarios of how my two early reporter suggestions, may have their complement or reporter end up pairing up in a design. And what stood out to me is that the core attractor sequence is directional. This is nothing new, however it has complications.
 
Meaning depending of weather the direct reporter complementary or the close to identical to aptamer sequence reporter suggestion I made is picked, they will not have the same options for aptamer turnoff. It will matter which is chosen for what pairings are possible and from which side of the aptamer.

I started wondering if there was a way around it. Because it will of cause being most optimal to leave the room open for us to target the aptamer for turnoff from most possible sides.

Then I recalled that the FMN aptamer actually had palindromic sequence inside it. AGGA and GAAG.

Actually all the aptamers mentioned for the coming lab, harbours palindrome sequences. Even more if pyrimidines or purines are counted instead individual bases. Just a quick and bad example is CGGU instead of CGGC or UGGU. Or PGGP (p=pyrimidine)

I have ended up thinking that palindromic sequence could be taken advantage of, as it would leave an extra option for which side

I have started wondering if not palindromic mirror sequence is second nature to aptamers and switching elements in general.


Directionality of the aptamer turnoff sequence


Here is my favorite Riboswitch on a chip lab, as it was the one that made me realize that we could choose what FMN region to target, just by reversing the FMN and MS2 turnoff sequence.

Perushevs design with palindromic sequences highlighted

 
http://eterna.cmu.edu/game/browse/5736148/?filter1_arg1=5764818&filter1_arg2=5764818&filter1...

I gave an intro to it here


One could simply choose what side of the two FMN aptamer sequences to aim for, by just by reversing the turnoff sequence. I described how to target an aptamer from a specific side.

Mods of Perushevs fine design ended up yielding the best scoring design in Exclusion 2, in the Riboswitch on a chip labs.

With palindromic parts highlighted

http://www.eternagame.org/game/browse/6369184/?filter1=Id&filter1_arg2=6456004&filter1_arg1=...
 
Note that the MS2 and FMN turnoff sequence has the palindromic sequence UCCU. One that the original design scoring 88%, didn't had. (it had UACCU)

Similar the Same State design with best fold change (when dumping a design with high fold change error) also is high in palindromic sequences - namely because it has a double FMN aptamer. Which in itself is double palindromic. Which the MS2 is too by the way.

Sara palindrome by Jennifer

http://www.eternagame.org/game/browse/6369196/?filter1=Id&filter1_arg2=6461353&filter1_arg1=6461353

Actually this palindrome thing could also explain the huge turnup of GUUG and UGGU sequences in switch labs also. CAAC and ACCA could also be a thing. :)



Alternative sequence suggestions in palindrome style


These still keep both the CC and GG magnets and in same position, but makes each halves its own palindrome.

1) Reporter with close to identical aptamer sequence

AGGAUCCU

This is really an abbreviated MS2


2) Reporter with close to aptamer complementary sequence

UCCUAGGA
(Edited)
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
It even hit me that the strands in the stem after the FMN in Mats winning mod of Perushevs original design, has palindromic nature. Only the highlighted G breaks the palindrome style of the left strand.


http://www.eternagame.org/game/browse/6369184/?filter1=Id&filter1_arg2=6456004&filter1_arg1=...
(Edited)
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
Here are two 10-base possible reporter sequences that holds all three palindromes of UCCU, UGGU, and GUUG by overlapping them:

AUCCUGGUUG


UCCUGGUUGA
Photo of Astromon

Astromon

  • 192 Posts
  • 26 Reply Likes
GCCG!
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
Some of the interesting things we learned from the FMN puzzles are that:
1. Players can indeed design 'perfect' switches, that perform as well as one could possibly expect.
2. The position of the aptamer and the reporters matter.
3. Limiting the number of bases that change base pairing between the ON and the OFF state may improve the switching.

Although we may have mentioned it in presentations etc., I wanted to show again an example of what we mean by perfect switches. The maximum fold change you would expect for the reporter binding is dependent on the free energy you provide by adding FMN. It also depends on the KD in the ON state, as shown by the red line in the figure below which represents the thermodynamic maximum (based on a simple 4 or 5-state model)

The 'optimality' of the switches is a good way to compare it with other natural or engineered switches that are typically far from perfect. It is for this reason that we would like to have a reporter and inputs that allow us to do this. For the OpenTB puzzle, we want to detect a ratio of inputs and it doesn't necessarily matter if it's reversible or 'perfect' as long as it works. For the switches (and papers), however, those qualities are really valuable. This is why I'd like to stay away from long reporters. Even though they may only bind with a subset of their bases, it is hard or impossible to know exactly how they bind and then calculate the predicted thermodynamic efficiency. In other words, we wouldn't be able to calculate the red curve in the graph above and not quantify the switching in relation to the free energy from the ligand. The left of the graph, where some of the best designs are, would be a grey zone with unreliable information.

For new reporter lengths and/or sequences we can calculate the predicted KD for a complementary target. It worked quite well for the 10-nt reporter we used in the past.

In terms of the placement of the aptamer arms, I think it would be of great interest to test explicitly whether surrounding the reporter by the two aptamer arms yields better results than placing the reporter next to the aptamer (e.g., Same State 2 NG vs Same State 1/3 NG).
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
@rhiju: If an oligo folds internally, NUPACK includes the calculation of the frequency of two instances of the same oligo binding to each other, e.g.
In this case is the formula for KD going to be
 ([free reporter][free comp][reporter•reporter][comp•comp])
    /[reporter•comp] 

or something else?
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
@Johan wrote:
In terms of the placement of the aptamer arms, I think it would be of great interest to test explicitly whether surrounding the reporter by the two aptamer arms yields better results than placing the reporter next to the aptamer (e.g., Same State 2 NG vs Same State 1/3 NG).
Hmm. In that case, for Theophylline and Arginine with "split" aptamers, we might want at least two different sub-labs with different numbers of bases between the two sides of the aptamer (small loop, big loop). This might help support different strategies for dealing with how to "occupy" the reporter binding site in the OFF case.

At a guess though, I would say the smallest loop or tail of variable bases should be 12-14 bases, giving enough for a reporter bonding site plus at least 2-4 bases to help switch it OFF when unused, or enough to create a strong hairpin to bind up unused bases.

Earlier I said:
The theophylline aptamer has fixed-8 bases on the 5' side and 11 fixed-bases on the 3' side, for 21 bases overall. If the working length for designs is still 85 bases, that leaves 64 bases to work with. If we want just one case, I might suggest splitting this fairly evenly as 21 bases for the tails and 22 bases for the internal free structure. If you end up insisting on a 3-pair stem on either end, that still leaves 19 base tails and 18 bases of free-form internal structure. It may be a bit tight but should be workable.
A small loop case could use 14 bases on the inner loop, with 50 left (25+25) for tails. A big loop case could have two 14+14 tails, leaving 36 for the inner loop.

Thoughts?
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
try to avoid strand1-strand1 and strand2-strand2 binding -- those could occur in the experiments and will complicate interpretation of the data (and eventual use in paper fluidics). 

but quickly, KD will still be:

[free reporter][free comp] / [reporter*comp]

(the other vals for, e.g. reporter-reporter don't enter the expression)
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
@rhiju OK. strand1-strand1 and strand2-strand2 bindings happen when the strands by themselves fold into a hairpin.  So we should avoid reporters that fold on their own.
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
@omei:
Would such designs not useful for the paper? Are the measurement errors associated with low KDON values such that the data from multiple designs can't be combined (via averages or medians) to provide meaningful conclusions?  OK, that was two questions, not one. :-)
I'll try to clarify. Unfortunately averages will not work. The problem is time. As you increase the length, both the kd and the off rate decrease. To reach equilibrium at concentrations below the KD, you have to then wait a longer time. We could, in other words, measure 10X lower KDs but we would have to wait 10x longer (many hours) for each concentration. The experiment would then take too long (things can fail), the RNA could start to degrade. We would also have many more binding sites than reporters in solution, making accurate measurements hard and requiring constant flow.
(Edited)
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
I have been thinking about how to get both a short reporter, one that doesn't have too many strong bases and one that allows us to make complements that can hook up with the aptamer sequences that I find of interest.

Omei mentioned that the long reporter in round 2 paired up in different ways with bulges and mismatches. He showed a set that don't start at the same spot.

Round 2 also was way more tolerable to GU's forming between reporter and reporter complement.

Many of the reporter complements actually hold double G magnets, which are made by adding a G forming a GU pair against the reporter.

I saw one design by AndrewKae that gave me an idea. This one holds two G magnets in its reporter complement.


http://www.eternagame.org/game/browse/7113220/?filter1=Id&filter1_arg2=7140036&filter1_arg1=...

This reporter holds capability of sticking most of the reporter complement I wish to make fitted to different aptamers in.

11 bases highlighted, demonstrating I can basically fit in different variants of the sequences I picked out earlier as interesting in against a stretch of the round 2 reporter.
 

This also keeps the reporter in more weak bases - something which should also help secure that it doesn't stick too tight and get too low a KDON.

And one can stick in the G and C magnets in both orders, as to fit the different aptamers better. So it holds a reverse option in itself. Also it carries elements of both my two early reporter suggestions, where one were close to complementary to the aptamer stretches I'm aiming for where the other was close to identical to the aptamer stretch I am interested in.

Plus the complement can even hold 3 palindrome sequences. :)

AGGA, UCCU, UGGU

So this reporter stretch really don't need to change, but we can make the changes to the complements to get us a bind both to reporter and aptamer as needed for the different aptamers. I think it is the best compromise I see for now between all the current demands.

Another advantage with keeping this original stretch of the round 2 reporter intact as is, is that we already have experience using it. 

Stretch I find interesting from the original round 2 reporter (11 nt)

GUCUUGAAUCA

I can live with loosing the first G or have it translated to a weaker base (eg U). Because one can always create the AGGA palindrome sequence in the reporter complement by prolonging the beginning of the reporter compliment with an A + GGA. Similar one could loose the A at the end for the same reason.

One potential problem with this shorter reporter and what I suggest is that the reporter may not work as well when short when using mismatches or GU basepairs against it. As the shorter 10 nt round 1 reporter was grumpy when it came to most mismatches and GU pairs.
(Edited)
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
Sequences:

MS2: ACAUGAGGAUCACCCAUGU
Tryptophan: CGAGGACCGGUACGGCCGCCACUCG
Theophylline: XCAUACCAGX|XCCCUUGGCAGGX
Arginine: GCAGCCGCGCGUGGCCUUGAUGUUGUGCCCGGGCUGC

Shared subsequences:

MS2,Try: GAGGA (The: AGGX, Arg: GGG)
MS2,The,Arg: CCC (Try: CUC)
Try,The,Arg: GGC (MS2: GG)

Similar subsequences:

RCYC: MS2=ACCC, The=XCCC, Arg=GCCC, Try=ACUC
RCC: MS2,Try,The=ACC, Arg=GCC
RGG: MS2,Try,The=AGG, Arg=GGG
RGGC: Try=CGGC, The,Arg: UGGC (MS2: GG)
YRCC: MS2=CACC,Try=CGCC,The=UACC, Arg=UGCC
(Edited)
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
any updates omei with 'correct' subsequence AGGU ?
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
Yes.  I've tried a lot; here is my favorite in terms of having comparable KD and self-binding.

Sequence: GUCAUAGGU KD=19.2

By shortening it to 9 bases, I was able to get a fourth strong base in.  Although I don't know that it will be better than 3/10 strong bases, my gut feel says it will be. 
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
@Omei wrote:
UACCUAGAUA has a predicted KD of 8.26, and seems like it should be in the running, with less than a factor of 2 lower KD than the legacy reporter.
The WC complement of this is UAUCUAGGUA which has "AGGU" one base removed from the 3' end.

You could also try AUAUCUAGGU if you want to shift it fully to the 3' end and still have all Y/R triples
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
Here's jandersonlee's nominations, in the same form

Sequence: UAUCUAGGUA KD=8.3


Sequence: AUAUCUAGGU KD=14.0


The reporter-reporter binding is somewhat stronger, but I don't know how significant that is.
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
AUAUCUAGGU KD=14.0 is very similar to the round 1 reporter KD of 14.3. If you want something that binds a bit tighter (some players complained the first round reporter was hard to bind with) then UAUCUAGGUA KD=8.38 seems fine too. GUCAUAGGU KD=19.2 is probably also workable, though it would be nice if we could try designing with it for a few days to a week before committing to it.

What if we run two (or three?) sets of labs with different reporters for a week or two and then have players vote on which lab they prefer to run with (freezing the other lab(s) without synthesizing them)? The labs usually run for 4-8 weeks lately, so taking one week to limber up shouldn't slow things down by much.
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
A question on lab tails: We no longer see them, but I assume they are still there. What do they look like these days?

If I recall, some early switch designs appeared to have trouble when they had subsequences that mimicked or paired with parts of the lab tail. These included GUUG (which paired with CAAC in the lab tail) and UUCG (which looked like part of the barcode hairpin) and UCC (which paired with the leading GGA).

Are the lab tails still the same as before?
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
I'm pretty sure the tails that were needed for the cloud lab experiments aren't present in Johan's experiments.  Some time ago Rhiju told me that Johan's RNA was tethered on both sides by DNA.  But it was not central to what we were discussing,  and I didn't ask for more details, such as whether it was single or double stranded, or whether they were directly or indirectly bound together.

I am particularly interested in the answer now because I find a strong indication in Johan's data that there is a thermodynamic bonus for having a reporter/complement stack form at exactly the 5' end of the design.  My first guess is that this is another manifestation of coaxial stacking -- this time between DNA and RNA helices.  So I'll be very interested to find out whether there really is an adjacent DNA stack.
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
@omei the reporter doesn't have to be 10 nucleotides -- it would be fine to make it shorter -- can we then get good pattern matching (like what jandersonlee & elifisker are suggesting)  to the targets but keep the Kd similar to the prior oligo and no self-structure? 
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
@omei: There are indeed flanking sequences on both ends of the RNA.
The construct looks something like this.

The design (NNN...NNN) is flanked by RNA on both ends. Both of these flanking segments are blocked by DNA oligos up to the base that precedes or follows the design. Stacking is a possibility. The flanking DNA oligos are quite long (31 and 39 bp) so they should be bound at all times.
(Edited)
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
@johana, @rhiju: Eli, Jeff and I have all been presuming that having a reporter that doesn't fold internally is a good thing.  But having a self-folding reporter would raise its KDON, wouldn't it?  

If we had a longer reporter (in order to give players more options for designing successful switches) but Rhiju's NUPACK test yielded a KD estimate comparable to the 10-base reporter, would that be a win-win choice? 
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
@Omei: Having a self-folding reporter may lower KDON, but it doesn't solve Johan's other complaint for switch labs: a longer reporter often forms a longer/tighter helix when it binds and is more reluctant to "let go" in order to change states.

As Johan also mentioned, that is possibly less of an issue in the OpenTB challenge for the actual in-field testing device, which only has to test *one* state, but for the labs being tested in multiple states it is an issue. We need to have the in-vitro tests "switch state" in a reasonable time-frame after changing the oglio concentrations.

We players could help with this by only targeting a maximal subset of 10-bases when the reporter was longer, which might give us more options on say a 14-base reporter which could then seem like 5-different 10-base reporters, depending on which run of 10-bases you targeted, but would probably drive Nando crazy trying to code up the labs, since only 4-bases (in the middle) would always be expected/required to bond; or not-even that if you allowed inner-loops and bulges.
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
I may be wrong, but my sense from Johan is that low KD is the most relevant issue for this project. Johan has modified his experimental protocol to make it much more robust wrt "unfolding".  (As well as take much longer.)  Because it would have taken even longer to actually measure for a slow unfolding effect, he didn't do that for R104. But after he had the data in hand, I asked him whether he thought all the various 20 base oligos and 14-base reporter together had adversely affected the measurements because of slow reversibility, and he said he thought they were probably OK.

Anyway, that's what prompted me to ask the question.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Stiffness of reporter and KDON


I have been wondering if something else is in play for KDON of reporter. And how tight it binds.

What I wonder is if a simple fact as if the reporter is being either (mostly) purine or pyrimidine will affect its KDON. I pretty much bet it will.

I have been saying that I think it is beneficial for an input/reporter to be biased in sequence. Because this is what I have seen for the microRNA's.

And since the input/reporter is to be an active switch element I would like to see it biased in either GA (purine) or CU (pyrimidine) direction.



Rigid design/floppy reporter or floppy reporter/rigid design?


I mentioned earlier that I thought purines were more rigid and pyrimidines were more floppy.

I simply think it will matter wheather it is the design or the reporter and inputs that are floppy. Each way around will not be equally good. At least not for everything.

Will the design be floppy if the inputs and reporter leans toward purine (rigid) or will the design be rigid if the inputs and reporter is more purine (more rigid)?

For round 2 the design was biased towards going CU, since the inputs were biased towards GA. (rigid inputs and floppy design)

Result for now: hard DEC design getting best global fold change

It was the opposite for round 1. Inputs were biased towards C and the design was biased towards G. (Floppy inputs and rigid design)

Result for now: hard INC lab design getting best global fold change

While it is still early, I wonder if these pyrimidine/purine base bias are affecting ON and OFF labs differently.



Protein analogy

I also found a protein anology for it mattering if the peptide (input/guest) was rigid compared to the protein (design/host)

My protein book "How proteins work" says that bound peptide will be low in entropy and fairly rigid when bound, but to be fairly rigid when bound, it should also be somewhat rigid already before binding to not loose too much entropy when binding. (page 150)
(Edited)
Photo of Astromon

Astromon

  • 192 Posts
  • 26 Reply Likes
Just want to point out that neither  purine or pyrimidine bases Bind with themselves.
(its just something i noticed when i looked the two words up)
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
Actually G and A can form a bond that is slightly wider than a GC or AU pairing. That is what provides a G "boost" to a loop. CU and UC are also slightly wider.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC134247/
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
@elnando888, @rhiju, @johan: Here's a what-if...

What if we had a puzzle-maker/lab-maker tool that would let players specify the reporter sequence? We have a *lot* of suggested reporter sequences, but no evidence/practice that they could be useful in creating lab designs.

I'm not suggesting to run the actual wet-lab round with a variable, user-specified reporter, just an in-silico pre-lab test-run to help "test-drive" various reporters across the different aptamers.

How difficult might that be to produce?
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
I totally get that Flash is dead (or undead/zombie). No, I would not want you spending weeks to make that kind of mod.

As I mentioned elsewhere a similar result could come from running multiple labs with different reporters for a week or two and then deciding which one to run with for the full wet lab -- but if you are "creating lab puzzles by hand" is that really any better?

I used to have a boss who said any new feature was a "small matter of programming" but he  never mentioned the time/complexity involved.
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
Creating lab puzzles by hand doesn't take weeks, so I'd say it would be slightly better, yes. One of the first thing I would need to do is to write a painter tool that can generate the complementary sequence of whatever oligo is used as reporter in these puzzles (just so I don't have N instances of such scripts for N different oligos). Then, I could rather easily make a container project for prototyping these aptamer+oligo puzzles.

Do we have a "top-runner" oligo yet? And how many other solid candidates do we have at this point? Could someone please summarize?
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
I see you appear to have already written this. Thanks!
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
Regarding OpenTB reporter oglios, the last input from Omei on KD evaluation was about two weeks ago and the main contenders seemed to be as follows:

Sequence: GUCAUAGGU KD=19.2
Sequence: AUAUCUAGGU KD=14.0
Sequence: UAUCUAGGUA KD=8.3

All compared to the to the round 1 reporter KD of 14.3.
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
I suggest using a lower Kd, i.e., 8.3 nM.

When I calculate the Kd, I get consistently lower values. I think that the NUPACK values depend on the input concentration of the complementary oligo, so I typically use a really low number, e.g., 1E-6 pM. For the 3rd sequence, using Matthews et al. 1999 parameters (advanced options) I find KD=0.7 nM.

The values for RNA in NUPACK are calculated using 1M NaCl and 0.0 MgCl. In the experiments, we use 80 mM KCl and 10 mM MgCl so the number will be different.

For the 10-nt reporter we have been using, I get 5 nM for RNA. For calculations with DNA for the same sequence (with Ts instead of Us) I get 106 nM for 1 M NaCl and 413 for 0.08 M Na and 0.01 M Mg. There is, in other words, a 4X difference due to salt.

This is consistent with our experiments, which show 7-20 nM for the reporter. This number is somewhat higher than the 5 nM calculated in NUPACK.
Photo of Brourd

Brourd

  • 452 Posts
  • 82 Reply Likes
I have a minor, silly question. So what would be wrong with using the legacy reporter sequence?
Photo of rhiju

rhiju, Researcher

  • 403 Posts
  • 123 Reply Likes
that would be fine with us. however in the recent openTB rounds, it seemed important to have player input in defining input/output strands -- we had successes in 'round 2' only after soliciting specific sequence suggestions.

does anyone see any issue with the legacy reporter sequence? And can someone (omei?) run all the proposed reporters and post here?

then we can hopefully come to a consensus by end of today (22 dec.) -- we have so far been delaying the puzzle setup.
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
Hope everyone who celebrates Christmas and/or Boxing Day had good ones!

We've decided to start off the next lab round by going back to the MS2 hairpin as the output, as we did for the FMN switches.  Nando is making the puzzles, but he is traveling and getting a connection can be hard. For now, you can test out the puzzles before they have gone live at http://www.eternagame.org/web/lab/7559749/.  If you discover any issues, post them here. 
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Overall aptamer/MS2 blueprint

This is amazing news. I simply can’t be happier. This is far better than a “mini MS2” just containing both magnet segments. This is a good choice for a lot of reasons. This will now be direct comparable with our past MS2/FMN experiment.


This is an especially good choice as this means we can practically reuse the MS2 and FMN positioning that worked best.


I here is a simplified version of the overall FMN/MS2 riboswitch blueprint I shared earlier - one that also takes into account that double aptamers worked also.


Quick redraw


I recommend reading this background post.


Using the MS2 is also good as it spare us the choice of picking if the reporter should be a negative to the aptamer or have sequence similarity. With the MS2 both choices stands open as it carries the CC and GG magnet segment and so do the aptamers. We can both aim to do direct pair up between MS2 and aptamer or use a middleman sequence.


Also I have done some thinking on the actual situation of using different FMN aptamers against MS2.


Sorry in advance, Johan, for us having MS2 ruining your chip...

(Edited)
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
I would like in particular to see the tryptophan aptamer split in two and placed at a position like theophylline in the Theophylline B puzzles. (at an FMN position as in the blueprint drawings above.) This will allow it to behave more like a FMN aptamer and the MS2 to take its favorite middle position.

I would also like to see something similar for the argenine puzzles. Despite argenine actually resembles MS2 more than an aptamer - meaning that it will likely favor a middle MS2 like position, if together with an aptamer like FMN. However when there are two MS2 like beasts, I will still like to see MS2 at a middle position also.

It may be harder to get hold of the argenine aptamer from the orientation it currently has. From the look of it it will like to be grabbed from the top, closest to the argenine molecule. So it may favor reverse orientation, when not given middle MS2 like position.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Making a great MS2/aptamer switch


There are certain things I think are particular beneficial for making effective riboswitches of the type MS2/aptamer.


Overall puzzle

Designs with double aptamer switches don't need static stems, by using repeat sequence and palindromes they up the chance for getting an extreme strong fold.
 

Aptamer

  • The best aptamers are splittable in half. (carrying an internal loop) Resembling an aptamer more than an MS2 hairpin.
  • Carry magnet segments either GG or CC
  • Carry palindrome sequence
  • Is structurally mirroring each other, even sided just like the FMN (symmetric aptamer) and not like the theophylline (asymmetric aptamer). This I say because I think symmetry is central for making a stronger switch.
(Edited)
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
@Eli: the Tryptophan aptamer cannot be split, well, at least not easily, so we won't be trying. [edit: we actually managed to do so, so Tryptophan exists now in both versions] As for the Arginine one, I'm waiting to hear back from the scientists, because I don't know exactly where and how to split the aptamer properly. As soon as I have the information, I'll add the two additional targets to the list.

It's probably obvious for most, but the labs aren't active, so please don't submit sequences yet.

And yeah, as Omei mentioned, it isn't trivial these days to pass through the Great Firewall of China, so apologies in advance for the unavoidable delays.
(Edited)
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
@Nando, big thx for making the Tryptophan variant and planning on the Arginine one. 
Photo of johana

johana, Researcher

  • 96 Posts
  • 45 Reply Likes
After a lot of deliberation, we have now settled on the aptamers to use. Nando has been great and updated the proposed puzzles (same URL as before: http://www.eternagame.org/web/lab/7559749/). It turns out that the original arginine aptamer was too weak. Nando suggested another one, with better affinity for arginine, and after comparing it with dozens of other aptamers it remains the best one. It's relatively small, we can split it, it has magnet segments, the ligand is easy to get and is soluble. It should be a great puzzle!

We will soon launch the MS2 version but would also like to create reporter oligo counterparts. The 10-nt is still on the table but I also liked the following suggestion above:
Sequence: UAUCUAGGUA KD=8.3
It's experimental KD is probably less than that but higher than 1 nM (0.7 nM in 1M salt). This is right in the sweet spot for experimental feasibility and large fold changes.

 @Eli @Nando @Omei @jandersonlee: Is this sequence still good with the new arginine aptamer?
Photo of jandersonlee

jandersonlee

  • 555 Posts
  • 130 Reply Likes
I'll defer to Eli in that matter.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
I would still very much like to have a CC magnet in there. Like UACCUAGGUA.

However I get from earlier posts that having both a CC and a GG will raise KDON too much.

I will get back on this after further thought.
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
The new arginine aptamers still do satisfy the criteria I used when I proposed using the subsequence AGGU in the reporter.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Ok, I think I can work with the reporter Johan suggests, as I will get my wanted CC magnet as a complement between the reporter sequence GG and the aptamer GG's.

Also the labs with reporter as a mini microRNA input + aptamers will behave differently to our usual MS2/FMN switches.
Photo of spvincent

spvincent

  • 49 Posts
  • 10 Reply Likes
When is this puzzle going to go live? There's this faintly ridiculous countdown timer (anyone would think you're about to launch a space shuttle) on the home page that has been started several times but it keeps expiring without so much as a comment. Bit of an anticlimax really.
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
I think they are ready to go, and it is only a miscommunication among the devs about who was going to make the project active.  I''m trying to find out.
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
I think I've managed to activate the first 12 targets (the ones with MS2 signal). That was a first for me, so I hope I didn't screw or forget anything important... Please let me know. 

On a related note, prototypes for 16 targets with oligo reporters are ready for review at http://www.eternagame.org/web/lab/7579422/ If possible, I'd like interested players to test them and provide some feedback before we conclude that they are indeed ready for launch, or possibly decide that we'd rather try a different oligo.
Photo of spvincent

spvincent

  • 49 Posts
  • 10 Reply Likes
Seems to be working tx although there's no announcement in the newsfeed.
Photo of eternacac

eternacac

  • 274 Posts
  • 19 Reply Likes
test of the Same State Arginine-A with reporter gave 3 pretty quick solves that are direct analogues of the current lab. One of them a Vienna+Vienn2 solve and only missed 1 box of NuPack (reporter was bound in state1). This one is ready to go, imo.
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
@eternacac : it would seem that we will be going for a different system. While solving these puzzles is certainly interesting and enjoyable, I would suggest at this point that you don't focus on these targets, which we possibly won't use.
Photo of Eli Fisker

Eli Fisker

  • 2239 Posts
  • 498 Reply Likes
Second state needs fewer basepairs than first state.

Spvincent asked a question for the current round lab that I wish to bring up here:

"For these switch puzzles, are there any guidelines as to what the difference in free energy between the states should be? I'm guessing this issue has been addressed before but I can't find any references."

Here is an indirect answer as this also effects the free energy difference between states. I wish to highlight something, that can be used to make or spot the better of the designs.


Rule of thumb for good riboswitches


The state with the aptamer molecule tend to have fewer switching basepairs than the state without.

Basically you can view the difference in basepairs between states as a way to push the switch in the direction you want it to switch. If there are too many switching base pairs in the first state it will not feel like switching to the second state and vice versa. The trick is to keep the switch not having too different amount of basepairs between the states, so it will feel like doing both states and switch between them.

As a rule of thumb the 2 state (with aptamer) will have fewer basepairs than the 1 state (without aptamer). This is the main trend for Riboswitches like Same state NG 2 and Exclusion NG2. (All our new B labs are  equivalent to those).

This is due to the second state having the aptamer. The aptamer molecule adds an energy input. In other words, the aptamer + molecule is helping holding the structure, such that this state need fewer basepairs for it to hold.

This is even more true for the Same State type of design, as it also have the MS2 hairpin binding the MS2 molecule in state 2, and also contributing energy, thus needing fewer basepairs to hold the structure.

The Exclusion type (OFF) of design though have lower difference in basepairs between states than the Same state type. (ON) This is due to the MS2 being present in state 1 and adding energy there.

Note: I have been viewing the designs in the Vienna2 energy model.


Labs with less optimal element placement

The trend for the state with the aptamer getting the fewest switching basepairs are not as profound in the other sublabs that have their inputs placed in a less optimal way for a solve. (Compared to the two best labs of each type - Same state ng 2 and Exclusion NG 2.

For some of the labs that have trouble reaching a winning score there is even a start with pattern reversal, so that among the few topscorers/winners are designs with the aptamer state holding more switching basepairs than the non aptamer state. (Less optimal)

This is even more true for Exclusion labs that carries less protection against reversal as they have the aptamer state fight with the MS2 state. Where the Same state labs have both MS2 and FMN in the 2 state, helping to hold the structure.


Background

I have earlier mentioned that there were differences in base pair counts across states in winning riboswitches. But also that switches that didn't have a huge difference in basepairs between the states, tended to do better.

 
Photo of spvincent

spvincent

  • 49 Posts
  • 10 Reply Likes
Tx Eli but I'm not sure I'm going to go to the trouble of counting base pairs: eterna is rather hard on my eyes as it is.

In any case this information is presumably encapsulated in the delta G values that are presented for each structure, and I was really interested in what would be considered an optimum value for the difference between the free energies of the two states (difference in differences to be picky about it I suppose). If the values are too close to one another there'll be no thermodynamic incentive to switch states, while if they're too far apart the RNA will get stuck in one state more or less irrespective of the ligand concentrations. I'm guessing a difference of 2 kCal/mole might be suitable here but am perfectly prepared to be told I'm wrong.

By way of background, here's an interesting read on just how narrow the energy range is for the chemical processes on which life depends.

http://wavefunction.fieldofscience.com/2016/03/the-only-equation-that-you-should-know.html#comment-f...
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
Just for the sake of mentioning it (I know it wasn't you @spvincent who initiated this), this particular topic is completely off-topic here. This thread is about the modeling of the labs themselves, not about design guidelines for candidate sequences. 

This said, I feel compelled to address your statement:
this information is presumably encapsulated in the delta G values that are presented for each structure
That's most certainly not the case. The Eterna interface provides us with MFE structures. These structures and their associated Gibbs free energies tell us very little about the partition functions (the ensemble of possible structures) of the considered puzzle states. In Eterna, we only have partition functions (the dotplot) for states without ligands. Provided you'd have both, you could possibly imagine balancing them by focusing on the particular structural motif that is used as signal. For MS2, try to minimize the probability of its appearance in one state and maximize it in the other. And that's exactly what I implemented in my bot ViennaUCT. But even so, you'll definitely run into another issue that partition functions cannot help you with: free energy kinetic traps. In other words, how hard/easy is it, to change from conformations to other ones? This type of free energy landscape analysis is possible, but extremely costly in computational resources, and given the quality of the folding engines, quite noisy as well (read: not fabulously reliable). In conclusion, the best we have at the moment (or at least, part of it) is possibly simple rules of thumb, the likes that Eli is proposing.
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 980 Posts
  • 308 Reply Likes
@spvincent The measured difference in free energy between states (ddg in Johan's spreadsheet) is simply the logarithm of the fold change.  Given a bunch of simplifying assumptions (like a perfect energy model, that there are only two possible foldings for the RNA, ...), the difference between the energies in the game would be the same as the reported ddg, and the larger the difference (while still folding properly), the higher the fold change.  But in fact, there is very low correlation between the two.  This was something we discovered and discussed early in the array-based labs.  That's basically the reason no one has stepped up with a definitive answer to your very reasonable question.
Photo of spvincent

spvincent

  • 49 Posts
  • 10 Reply Likes
Thanks for the clarifications Omei/Nando: looks like I have more reading to do!