Challenge puzzles version 2

Mat and I were discussing how to make the change from challenge puzzles to the lab puzzles a bit easier for the newcomers.

The rules for what work in the challenge puzzles are different from what works in the lab puzzles. How you could solve a challenge puzzle, is not necessarily the same that will give you a good synthesis results.

As we are discovering the rules for lab puzzles, I’m wondering if some of that rules (the ones we are most sure about) could be fed back in when making new challenge puzzles, sort of challenge puzzles version 2. Some puzzles specificly aimed at improving the players lab contribution.

Then the newcomers to the game could be feed those ”labchallengepuzzles" first, as they would help them do well in the lab. That could ease the lab experience for the new players and let them do overall better designs, so they then could worry about those rules/patterns we haven’t figured out yet. And when they feel like it, they can always go back and do the ”old” challenge puzzles as well.


We have been discussing this idea as well - lab tutorial puzzles that’ll make the transition from the Challenges to the Lab easier.

The tricky part is what to include or not in these puzzles. We have many player discoveries/theories - but not all of them have been proved to be correct. Do we want to plant all player theories to the new players nonetheless? or should we try to use ones that is absolutely provable?

I would love to hear player thoughts on this - or even better, ideas on what these specific puzzles will be!


I fully agree on that it will be tricky what to include or not. I was thinking only about, sort of teach the things we are most sure about. Else we might end up confusing the new ones more and even worse teach them halfwrong stuff, that might prevent them from seing the ”light” and new original problem solutions.

I think about things like these (that I am sure about:) )

Use wax with strong hold when you set the hair of your RNApuzzle: Why it generally a good idea to close the stems connected to the multiloop with the strong GC-pair. (The tetraloops as well, though in some puzzles, typical those with long arms, a closing with a AU-pair or even a GU one might be tolerated.)

Ring race in the loops, drive around in the same direction: That the sameturning GC-pairs in multiloops, with red nucleotide to the right, is usually a good idea. And that in designs with long arms, this rule can be broken.

Some twist it puzzles. Not using more than max two samecolored nucleotides in a row. That it’s generally a good idea to twist nucleotides, so the neighbouring ones have different colors. At the same time teach that longarmed designs are less rulesensitive, so in those it might be posible to get away with it there. They should know that it’s a 80-90% rule, just so they don’t rule out a possible solution in the puzzletypes where it is alowed. So puzzles with arms of different length could be used to show the difference in rulesensitivity.

Avoid making a christmas tree/cub scout: Maybe include a pair of puzzles with limited amount of GC- and AU-pairs usable, so it is not possible to make a pure christmas tree and the cub scouts, then at the same time tell why the one or the other might not be a good idea. Then again we don’t want to censor what people are designing, just guide.

The easiest solution would propably be that some of the other players and me each write a short lab guide with do’s and don’ts and specificly aimed at the new players. And preferable, as I remember that someone else suggested in the forum, place the guides together, so the new players can find them.

There could also be a textbox along with these lab learning puzzles with an explanation about what the players are about to learn, and why things is so and in which cases things is so.

Eli and Jee: This is a great discussion.

Attracting novice players has been much on our mind recently. We’re worried that, ironically, EteRNA is becoming too much of an “expert’s community,” and that newcomers are becoming intimidated.

A suggestion : Maybe interested players could design a set of puzzles using the puzzle creator. Then we would form a committee to decide to include them as easy challenge puzzles?

Thanks, Adrien.

I think your suggestion is great. It is not always possible to see from the numbers of stars a puzzle has recieved, if it is easy. Because two or three really good players come along and vote it as easy - because it is for them. Not many players in general vote on the degree of dificulty, which makes the result biased. Now we sort of have to go backwards (click back in the browser) to vote on a puzzle when we have finished it. Could we instead be asked to vote after finishing the puzzle? (Still as a choice, not an obligation) Then I think more players would remember to vote. The problem today is that too few vote. When I have finished the puzzle, I think about the next and forget go back and vote. People primarely vote if the puzzle is really easy or really hard. I usally only vote when I find a puzzle hard.

I like the idea of a commitee. Maybe some of the recently new players should be in it as well, to tell us if the puzzles chosen really is easy.

Great point Eli:

I would point out that while it’s very difficult to say for sure what works, it a little easier to point out what doesn’t work, using previous lab results whenever possible so players can get a sense of what more experienced players already know just from observation, trial and error.

Puzzle players almost universally want to start with an all GU, all AU, or all GC designs, and just eliminating those from the starting pool would great in and of itself - even though they are all sure-fire puzzle winning strategies.

Likewise, learning to read the dot-plot immediately remedies the problem where highly repetitive, symmetrical designs are used because they also have a high probability of forming alternate structures (of particular note since all the lab target designs have been very symmetrical).

Hi Alan

Thanks for your comments.

You are absolutely right. There will probably be more don’ts than do’s, in my comming labguide.

And yes, dot plots are very useful, I forgot about them. This is why it will be great if a few of us make a lab guide. Each will remember something the others have forgotten about - that is seen with the eyes of a newbie. If you haven’t planned to do one, no pressure though - you have already made a useful collection of articles - I will just mention that Alan says… and tell them about dot plots as well.

I had great help from the two pages on your player profile, to understand how to use Viennas RNA-fold server. (RNA-fold and positional entropy and Kinetics and sub-optimal folds.) They should be on a collected page with resources for the beginners too. Okay here I admit the advanced beginners. :slight_smile:

If we shift to a lab selection scheme based on binary ranking or predicting EteRNA score, we could give player’s a few training puzzles in the same format. For example, we could minimally get players to recognize that Christmas trees are problems, or that designs with ‘bad dot plots’ are problems. [is that true?] Can you help us make a list of ‘inarguable’ but important lessons from the lab so far?


Do I understand you correct: That you would like to make a training puzzle and then let people solve it different ways, so they can get the fealing about what works or not – from the estimated score they get? That would be absolutely cool.

We could actually ask them to make :slight_smile: christmas trees and cub scouts, to let them see for themselves that this is propably not be a good idea. That could give them the feeling of the experience we already have. And keep most of the christmas threes in the playpen.

The only problem is that some of the christmas trees actually score fine – it’s just that they don’t have the potential to become the winner, which you can’t see from the score alone. So you might wanna go for the binary ranking, if that is some sort of ranking for which puzzles are good compared to the christmas trees and cub scouts. Or present the puzzle score with a text, explaing why this puzzle which scores great might not be the winning bet.

I would love to help making a list of ‘inarguable’ but important lessons from the lab so far. I have made a list of what comes to my mind first. Then you guys can select which you believe in and want to use for this practice puzzle.

Partly arguable but important lessons from the lab so far

Don’t make the arms totally identical. That will highten the chance that the structure will try to form alternate structures as Alan.Robot says. So they will pair up with each other and thereby mispair.

Don’t make the arms too different. Make the arms in you design in the same ‘mood’ of colors. Not a design where one arm leave an impression of blue, the next red and third green. Arms with very different collors means not just imbalance in colors, but also an energetic imbalance. The winning designs so far have the energy nicely spread out in the design, an almost even energy distribution.

Don’t use blue nucleotides in a multiloops, they have an ugly habit of breaking up things. Because Mat says so and I have demonstrated that they raise the level of energy inside the multiloop dramatically in positive direction, compared to nucleotides of other colors and thereby creating an energy jump.

And as Alan.Robot sugested it’s important the new players understands dot plots. Would it be possible to teach this along with the lab puzzle? That a messy dot plot means that some things have to be done differently. That a clean dot plot does not count, if the design is leaning towards christmas three. Tell them though that they don’t have to make a totally clean dot plot at first try, as failure to do so, could scare them away. Probably only a few of us did that anyway.

While using only the strong GC-pair all over the design is not a good idea, you should instead use your GC-pairs where their strenght is needed most, mainly in the junctions – like where a stack meet a multiloop and more often than not, when a basepair is closing a loop.

Tendency in the lab for tetraloops, is that they should have same value of energy inside (in a design with 4 tetraloops they should either be identical or should have a twin somewhere with the same energy level inside. 2 x2 identical.

Turn the GC-pairs in the multiloops in the same direction, red nucleotides to the right, as to raise the negative level of energy inside, to avoid a potential energy gap between inside multiloop and the arms connected to it.

The higher negative energy inside a multiloop – the better, as long as it gets the energy level in it more in line with the energy level around it. Which means don’t put colored nucleotides in the loop ring, unless they are green and raise the negative level, or red and you can afford to loose a bit of negative energy. And there is a right and wrong side to do this. (viewable in the energy tool)

And if no one end up arguing with my theory about energy jumps as a cause of mispairing, you might wanna include something about not making too big energy gaps, that is in stacks and between stacks and multiloops.

Make versions of the puzzle with different features automatically turned on. (The players might not have turned them on themselves yet.) Like putting the advance energy view tool in action, when teaching about energy gaps.

I might return with more suggestions here, when they comes to mind.

I would change the selection system in the lab so it is more appealing to new comers. Right now it seems to me to be vary click-ish. And I suspect that it is currently not selecting for the best design. I don’t think you need anymore training between the tutorials and the puzzles. Just more training on the tool they have like the dot pattern and the energy. I would not worry about the xmas tree problem I would add some negative points for poor designs which will solve the holiday problem. Anything below 70% synthesized gets -1000 for the designer and -500 for the voter.

I agree with Joshua on a couple of his points. I’d like to think what we learned in the puzzle series will be of benefit as we move on to more complex shapes in the lab. To date, the shapes have been extremely easy to create - the challenge has been in optimizing design for synthesizing. While GU challenges are fun, they have very little benefit in the lab.

As a relatively recent newbie, the dot plot was a challenge. I, not very long ago, posted that I was impressed with folks that could state the numeric location of a base pair they were addressing. Big DUH! As I was looking over some a few minutes ago, the light came on. It’s a bit of a pain, but now I can count the sequence and go exactly to the pair creating interference on the plot. Explaining that process to a lab new comer would really help to alleviate frustration and lead to better content. For instance, there is a design getting quite a few votes in the current lab that really doesn’t appear to be able to be synthesized well at all.

That leads to another of Joshua’s points - the voting system. Are you still planniing on rolling out a comparative voting model, i.e. selecting the best of two designs? If so, I agree with Joshua that the designer’s name not be included.

In any event, I would love for the community to figure out a way that more designs could be synthesized, even if it takes longer to get results.


Yes, more tool training for the new ones. That would help.


Good points about your travelsome way to be able to use the dot plot for more specific action. It took me a couple of months to figure out what the blue-yellow mode with paired and mispaired could be used for. :slight_smile:

Yes. We are planning on coming out with a new system. The “problem” is that the EteRNA community has been performing so incredibly well (even with the current clique-ish system) that we are reluctant to contaminate the experiment by changing voting system until we are certain to have all the data from this particular “experiment” (i.e. EteRNA v1, itself).

So this is not really a “problem” at all, but really a huge congratulations to and appreciation for the whole community.

But the community’s suggestions are constantly on our mind, and the new voting system is in the works. Jee has already showed us a working prototype.

While I’m at it, big thanks to Jee (the grad student leading EteRNA) for his amazing work.