The Strategy Guide to solve EteRNA Puzzles

My pattern

1
I paint one side and all loops in red, with preference for outside.
I paint the other side in blue.
I put an AU at the crossing between every stack and loop with blue on the other side than the stack except for little symetrical loops in the middle of a stack.
I put GC only at wierd angles.
At this point, often most of the stacks are whites with problems in angles and loops.
2
If a stack is not white, I put all it in the other side or I put an AU in the middle always in the reverse side , if it dont works at random.
So stacks are like this
AU or GC depending on the angle
UG
UG
AU
UG
UG
UG
UG
AU or GC depending on the angle
3
For loops, I put just the yellow needed to minimize energy never blue or green.
4
I move the GC until all is white and conditions requested are met : no strategy for this step
5
I try to replace every AU by one UG to have a place in competition

Very Interesting Variation on Matt’s popular method, Marie! - It sounds to me like it would eliminate much of the clicking on same bases characteristic of Matt’s method, and get you to a solution sooner & quicker. It seems to skip - or combine - the first steps and go straight to where you would eventually end up any way - Very much looking forward to trying your method. Thanks for posting it,

-d9

Quick Start with Eterna

Here’s my own mutated variation on Sneh’s original method:

  1. From the all-yellow initial state in target view, do a walk around the structure adding blue where ever there is an unpaired yellow. Do not bother making this intentionally random, just run blue bases up one side of every stack.
  2. Next, examine the little image in the top left corner that shows a summary of your current solution where red indicates bad sections and white indicates good sections.
  3. Using the Swap Bases tool (the 5 key), begin flipping yellow-blue pairs while watching the solution summary for the global impact of each flip. This is the phase I spend most of my time in. My goal is to get as much of the summary to the white state just by performing base flips with only yellow and blue bases. I think of this phase like lock-picking in reverse. Each time I flip a base, it’s like adding another notch in a key that prevents it from fitting other parts of the puzzle. I’m trying to jigger the bases so that, for the most part, they cannot line up in any way other than the target solution. Little changes can have a big impact and it pays to experiment all around.
  4. When I have ceased to make further progress with base-flipping, I toggle between target and natural mode. I note places that can only be fixed by adding red-green pairs. These are areas that line up properly in natural mode, but still lack the necessary bond strength to form the desired loops and bulges. I add the minimum number of red-green pairs necessary to lock down the structure. I don’t add red-green pairs in areas that are already all white in the solution summary at the top.
  5. If the puzzle has a minimum red-blue pair count, then I replace yellow bases in yellow-blue pairs with red in order to satisfy this requirement. Since these are the weakest bonds, one must avoid inserting too many of them in a row, lest the structure become too weak to hold together. Usually, I can get at least 2 in a row, sometimes 3. Note that this method starts by flooding the puzzle with intermediate bonds, then inserts the strongest bonds, and finishes by adding the weakest bonds.
  6. For very tight loops that refuse to stay closed, I may add extra unpaired red bases, but this seems to be a rare occurrence in the puzzles.
  7. There are also situations where it is only necessary to flip a red-green or yellow-blue pair to get everything settled, and it is worth the time to probe for these trivial solutions. Flipping pairs is a deceptively powerful way to tweak a structure, and it is also quite interactive since the operation is its own inverse (as long as you are in target mode, a second click will easily undo your change).

I just used this strategy to solve a level 4 puzzle. It works.

So, with minor differences, we have a pretty effective common algorithm.

But then I don’t really see the point of the game (he said, after spending ten hours playing with it!).

On the one hand, if there were a point to hand-solving, I might ask for this or that additional feature. On the other hand, it would seem a modest program could churn out a dozen ranked solutions in a couple of seconds.

In particular, it seems an AU painter feature would be nice, or even a better replacement for an all-A starting point.

I suppose I also remain unclear what the point of the synthesis lab is, compared to the game. Yes the folding algorithm needs validation, but how does pattern X relate to whatever the physical validation reveals?

FInally, my two cents to the algorithms, perhaps known to those trained in the field but new to me, was that swapping the AU pairs sometimes does make a difference, but the news is that most of the time it does not! That allows the easier painting of blues on the starting patterns and generally allows the random lengths of AU alignments to work.

Clean Cut Strategy Guide to Solving Puzzles:

#1: When starting a puzzle, determine if we are limited by the number of Red-Green pairs. If we are not, color every single paired match in target mode on one side first, with all red, or all green (but only ones that are side by side). If we are limited by the number of Red-Green pairs, color a single side Blue (but only ones that are side by side) so we have a Blue-Yellow lane.

#2: Begin flipping ONLY those pairs that are at the edges of their “lane”, i.e. before a bubble/circle and pay attention to the amount of redness in the puzzle’s thumbnail at the top of your screen. Play around with flipping them a bit, until the entire puzzle turns white. If we are limited by the number of Red-Green pairs, and we have gotten as much white as we can from the Blue-Yellow pairs, choose to place these Red-Green pairs ONLY at the edges of circles or bubbles. Consider the Red-Green pairs as “pinchers” in your lane, that create these bubbles for you.

#3: When we need to create Red-Blue pairs to fulfill the puzzle objective, if we are not limited by the number of Red-Green pairs, this means you will only need to turn the green dots blue. You can typically select green dots at a minimum of 3 dots away from the edge of any lane, without an issue. If there is a problem, typically flipping ONE pair at the beginning of any lane will allow you to begin at the second dot, instead of the third, and fill in many more Red-Blue pairs than you would normally be able to find without issue. If we are limited by the number of Red-Green pairs, typically you can use the same tactic with the Blue-Yellow pairs.

Following this, you can typically solve any puzzle in under a minute. I’m currently on the 4th page of puzzles, and still blowing through them in under a minute using this. If you do end up running into a problem, re-evaluate where you have placed your limited Red-Green pairs, and try flipping more Blue-Yellow lane edge pairs.

Very simple. To leave the tricks for the bottom:

Hexagon shaped dot pairs will need to have either Red-Green / Red-Red / Red-Green, or if we are limited, will need to have Blue-Yellow / Red-Red / Blue-Yellow.

Pentagon shaped dot pairs will ALWAYS need to have two pairs of Red-Green, with the point/top dot of the Pentagon left Yellow.

ALWAYS give the Pentagon shaped pairs your Red-Green dots first, with Hexagon and larger circles second.

So, basically 4 simple, repeatable steps. Energy has no bearing in these patterns. It’s simple force of attraction, and we know that Red-Green is more powerful than Blue-Yellow. Use accordingly.

This reply was created from a merged topic originally titled
Clean Cut PATTERN guide to solving puzzles.

After reading @korsen, "Pentagon shaped dot pairs will ALWAYS need to have two pairs of Red-Green, with the point/top dot of the Pentagon left Yellow. ", I reviewed my handful of Level 4 solutions. I found a few pentagons with GU and AU pairs matched with one GC pair and one where no GC pairs were required (try Xenopus tropicalis).

It took more than one minute to solve these puzzles, but my goal was to push as high as possible in the level 4 competition rankings, assuming that gaining skill at placing difficult GU pairs would be rewarded in the lab design and synthesis phases (TBD). I’ve gleaned from some of the staff and more experienced player postings that while extreme designs may quickly solve a puzzle, they won’t likely fold during synthesis (see "RNA Lab: “Christmas Trees,” “Cub Scout Projects,” and “Optical Illusions”). For me, having one of my designs gain high marks when synthesized would be the ultimate prize and of course the points would be nice too!

I think I agree with Aviel Menter. I thought the whole point of this was to have people reinventing the wheel. I did inquire about calorie management, but I knew that I could go through a tedious trial and error that would get me to the same results.
I enjoy thinking about the structure of uracil and the ways it makes RNA so different from DNA. I think that too much explicit guiding can prevent breakthroughs.

I like to go back to my solution and then try to tweak it to get a lower energy number. Basically the solution works but after I have something that works I want a chance to optimize it before submitting it. Granted I’m only doing the puzzle challenges at the moment. I figured I need to get a lot of practice there before trying one of the labs.

Thank you! Now that the puzzles do let you do this, I wouldn’t want to live without it.

Other strategy guides to puzzle solving have been made in the meantime and since they are not mentioned here, I thought I would add them.

Puzzle solving guide

Brourd’s Loop boosting guide

Jennifer Pearl’s Challenge Puzzles Quick Solve guide

Brourd’s Puzzle solving guide

Hoglahoo’s Reverse Spore Walk-through

Drake’s Advanced puzzle solving guide

Switch puzzle guide

Also check my PROFILE. There I will have an updated list.

Solution Algorithm:

Premise:
Considering that the strand starts its fold from a perfectly straight orientation, the bases are going to be drawn to pairs very close on the stack, so start at the hairpins to setup the target shape, working towards the ends last.

Drawbacks:
This algorithm assumes the strand starts with all A’s. I suspect that this default for all the challenges is skewing the solutions. I’d love to see some variation of starting stack.
This solution doesn’t minimize energy, but DOES focus on achieving the desired shape with the strongest base pairs.

Step 1:
Place G-C pairs at the neck of each hairpin, enough to start the strand folding the right way. Especially if you are limited on the number of G-C pairs allowed, this is the best mechanical use of these strongest pairs. Use up the allowed number of G-C pairs at the beginning of each new zipper.

Step 2:
Continue the zippers with A-U pairs, the second strongest.

Step 3:
When a zipper is interrupted by a loop, resulting in an angle, where the zipper resumes it probably needs a strong G-C pair up front to fold correctly, perhaps with the addition of the side G. If the fold occurs at the wrong angle, place a C base to prevent an undesired A-U.

Step 4:
If the fold is incorrect at this point, work from the natural fold view to interrupt undesired pairs by changing the bases in these regions, then fixup in the target view.

Step 5:
Now that the desired shape is achieved, add in the required number of weak G-U pairs. Three places they can go without ruining the target shape:
5.1: right after the strong G-C’s at the neck of a hairpin.
5.2: between stronger pairs at the middle of a long zipper. The longer the zipper, the more weak G-U’s it can hold without losing shape.
5.3: right BEFORE a zipper splits into a loop.

Optional energy reducing step:
Once desired shape is achieved, randomly swap out bases in loops to reduce loop energy, randomly flip base pair orientations to reduce energy, randomly swap out one base pair for another to reduce energy.

Just want to highlight that there now is a more complete LIST OF GUIDES in the WIKI. Check it out.

Thank you for taking the time to write this guide!

I have noticed that in places like

you would think that better is lower energy ( 2.5 ) but actually

better is 2.9.

But not like this

interesting… maybe it’s because another cross-molecule connection is, thus, not possible, and therefore stabilizing this molecule?

jrstern, For further improvements on algorithms used for MS2 / FMN labs, I propose the following:

  1. Leaving a long set of A’s (I will label these as “stretches of desert”) untouched may result in desired shapes. Such deserts simply won’t bind well (except maybe to the few U’s strewn around). However, the overall score of all these molecules is in the lower midrange of all the submissions analyzed here. So, leaving a maximum of 5 A’s in a row is desirable, if not even less.

  2. Maximizing the energy difference between the states seems to result in higher scores, as the FMN seems to really bind well, even if the target shape does not coincide (even remotely) with the predetermined binding site. Ultimately, all lab submissions that were created using this strategy had a vastly different shape than the one required, but all score better than average, with a maximum of 80/100.

  3. However (and thus already limiting the last finding), high fold-change in Kd,obs alone also does not necessarily result in a good binding site. For high Tmelt=107 °C or higher, low resulting scores seem to be the norm. Another rule to maximize the score therefore seems to be to keep the melting point between 67 and 87 °C.

  4. Geometric arrangements of non-A’s (for example an A or a G every 3 bases, aranged regularly) seems to result in very low folding score and for some reason no differentiation between state energies. So, another conclusion seems to be to avoid regular arrangements of blocker (i.e. non-A) bases.

Any thoughts are welcome.

This is what I have discovered from doing many puzzles:

Do not use mainly AUs or GCs. The way I see it is that both AUs and GCs are “strengthening” pairs and they shouldn’t be used unless necessary. AUs are great basic strengthening pairs when the GCs are for the “heavy duty” use only.

Fill your puzzle with GUs first, boost the loops and try to strengthen the structure with AUs first. Only if AU - GU combinations aren’t strong enough you should start putting GCs in the game. Places where GCs are most likely necessary are in the edges of big / main bubbles and small bends where there is one A between two pairs.

Best places for the AUs seem the be one pair off from start of the chain and / or in the middle of the chain (no always the case but usually yes). For example six pairs long combinations could look like this: “GU-AU-GU-GU-AU-GU” or if you need strong connections, “GC-GU-AU-AU-GU-GC”. Sometimes chains need to be built only from AUs or GCs but they are usually very short chains or in some very tricky places.

What seems also to make huge difference to the structure is that which way the pairs are placed. Flipping one or more chains might solve the puzzle without making any other changes to the puzzle. Flipping is usually the first thing I try before starting to place any GCs. Sometimes flipping long chains of GU pairs within the chain seems to also help and you might not need to use AUs or GCs at all.

You should also always keep the energy levels visible as the energy levels help you to picture more easily why some parts of the puzzle won’t work. Usually lower the number the better but not always. Sometimes higher numbers in the middle of the long chains or weakening of some parts might make other parts stronger solving the puzzle in best cases.

With this approach I can solve most puzzles in less than 5 minutes including the harder ones and have great amounts of GUs at the end. I simply don’t have so much success with AU only zipping or spamming the GCs. GCs also seem to have bad habit of messing the RNA structure which makes puzzle solving even more difficult.

You should try this next time you play and check if it helps solving the puzzles.

Great compilation of startegies, very usefull !