Down with GC!

  • 4
  • Problem
  • Updated 8 years ago
As dimension9 already said, as all the people who repeat it in the chatbox, there is a need to instruct players in a better way about why CG bonds are not the solution to everything.

I found d9's post on the repetition of the same bonds over and over truly enlightening, and that's how I came to understand the concept for the first time. That post, or the overall message at least, needs to be put right in front of players. Otherwise we'll keep getting christmas trees in the top slots of the lab (as we already did and are on the way to do again for the bulged cross part three). To do this, I have a few ideas:

1) Better tutorials:
Currently tutorials do a good job when explaining how things work, but they're not complete: there needs to be a tutorial about repetition, free energy, MP(the latter being a mistery to a great deal of player, me included) and loop and stack optimization.

2) Giving lab workers the possibility to discard a design:
Democracy is great and everything, but as everyone can see it has its downsides, one of which is that people not aware of the reason why christmas trees are bad and with a general understanding of how energy works WILL vote designs chock full of GCs. Giving lab workers the power to discard designs which will never bond as they're supposed to (like the B design in the Bulged Cross part 2) is one way to solve this problem.
Those people know what they're doing, while the majority of us don't really grasp how RNA really works. My best bet is that they probably won't go mad with power.

3) Put important articles (the ones on mpb21's profile, for example) on the front page, with a shiny button that catches the attention of the new player. Not everyone likes to read, but for those who do those guides are a lot helpful. I know there's plenty of challenges I wouldn't have done without the loop guide.

4)As already suggested(not by me), tiered lab voting would be great!

5)Dividing challenges into tiers. As in, to unlock level 2 challenges players have to complete (just saying) half or more of the level 1 ones. This way new players will need to complete more challenges to unlock the lab, and be more rna-savvy when those 10k come around.

If I come up with any other ideas about GCs I'll post them here, but that's it for now! Have fun and have a great day!

Christian
Photo of Christian Fratta

Christian Fratta

  • 5 Posts
  • 0 Reply Likes
  • confident

Posted 9 years ago

  • 4
Photo of Christian Fratta

Christian Fratta

  • 5 Posts
  • 0 Reply Likes
seeing how there are now TWO christmas trees in the top eight, bump bump bump
Photo of kyrstymoon

kyrstymoon

  • 3 Posts
  • 2 Reply Likes
I, personally, would really like to see some sort of quiz before you can vote in and design for the lab. Not technical things, just things like asking why lowest energy is not necessarily best, why all GC is bad (I really like the sticky glue and paper example), how to lower the energy of loops, etc. Things that if you've paid attention, read some of the articles and actually put in some effort you would know.

People might get the answers from their friends, but hopefully some people would at least read through the questions and realize that Christmas Trees are not a good thing! Maybe, *maybe* it would improve the quality of the designs we see?

I do think it's a good idea for lab workers to reject designs, though. There is just no reason to work on something that in reality will never bond properly.

As a side note, I've seen far too many times in chat where people say, "I'll vote for your design if you vote for mine!" No merit, no review, just blind voting. I know there isn't really anything that can be done about it, but it doesn't change the fact that it irritates the hell out of me.
Photo of jdbakermn

jdbakermn

  • 4 Posts
  • 1 Reply Like
I think there's some improvement to be made in the tutorials to teach lessons, instead of just teaching how to play the game. It seems like people (myself included) come out of the tutorials with some strange ideas (like lowest kCal is awesome, or that clearly the most GU bonds must be something really good) and then apply those ideas to the lab. I learned a lot outside of the tutorials (and challenges) by reading posts and looking at old labs, but only after I saw messages in the chats that led me to question whether I knew what I was doing.

A quiz to enter the lab might be a quick fix, but the tutorials need to be fixed too. Now they teach people how to play the game. But that doesn't prepare them to participate in the lab, as evidenced both by some of the submissions we've seen, as well as the "If you vote for me, I'll vote for you" attitude that KyrstyMoon mentioned above.
Photo of kyrstymoon

kyrstymoon

  • 3 Posts
  • 2 Reply Likes
That's a good point, the tutorials really do need to be fixed. A lot of the information people need to have could most certainly come from there. I remember being a little frustrated when I found out about adding reds to loops and how it lowered the energy, wondering why this information wasn't given up front.
Photo of Fomeister

Fomeister

  • 12 Posts
  • 4 Reply Likes
The best performing result to date (the all 33 pair GC not withstanding) has been the one with the most GC bonds, and ZERO GU bonds.

This is a complex project and it is entirely un-helpful for people to be drawing conclusions as to methodology.
Photo of Chris Cunningham [ccccc]

Chris Cunningham [ccccc]

  • 97 Posts
  • 13 Reply Likes
Personally I feel that posting static tutorials or Official Guides is a dangerous idea, since (as people like FoMeister have pointed out) we haven't really done enough tests to be sure that various things won't work. I wrote the paper/glue analogy, and I am no expert. So, I agree with you guys that there is a problem, but I hesitate to "hard-code" the solution (i.e. Tutorial #6, don't use GC bonds, here is why) -- I'd rather upgrade the ability of the community to share ideas with each other.

In my opinion the best way to do it would be a discussion thread for each submitted design. I'd like to not just vote for something, but tell the people why I did. Or, when I look at a design and I pass it up, I'd like to say why I passed. When I sort the lab submissions by #GC bonds, there are about 40 of them with all GC bonds. A kind-hearted person could gently point these people in the direction of the paper-glue analogy and the link to the failed design of round 2 and say "My theory (which comes from the following data) is that too many GC bonds makes an RNA fold the wrong way, so I didn't vote on your submission. Welcome to eteRNA though!"

At the moment, I can look at old lab results and learn a few things, and I can post threads on GetSatisfaction hoping that other people will tell me what they learned, but people who make a submission and get no votes don't have any way to get feedback other than spamming the chat asking for it. We need another way.
Photo of xmbrst

xmbrst

  • 13 Posts
  • 3 Reply Likes
Discussion threads for designs would be great, but it seems like there should also be a place to discuss the general theory of design.

To really discuss open questions about what makes a good design, you'd need to be able to discuss several examples at once. Without a place for general theoretical discussion, the threads attached to particular designs would make for a very fragmented debate.
Photo of Chris Cunningham [ccccc]

Chris Cunningham [ccccc]

  • 97 Posts
  • 13 Reply Likes
I kind of think GetSatisfaction might be working well for general-theory discussion? Maybe? I'm not sure.
Photo of xmbrst

xmbrst

  • 13 Posts
  • 3 Reply Likes
Can you recommend some good topics in that vein? I've found it hard to find such discussions, since the topics are all lumped into one big pile.

The "strategy guides" are easy enough to find, but are there places where people are talking about open questions -- things that are puzzling about why some designs synthesized well and others didn't?
Photo of Jeehyung Lee

Jeehyung Lee, Alum

  • 708 Posts
  • 94 Reply Likes
In this week's dev meeting, we decided to attack this problem in 2 ways

1) Advanced tutorials that will tell people why lowest energy is not the best, and some additional features that might be interesting

2) As ccccc said, we'll create a thread for each lab puzzle, and always put the link in the lab browser such that people can easily access the thread whenever they are reviewing designs.

These are makred as case #60 and #328 in our task list.

Thanks for great ideas!

EteRNA team
Photo of xmbrst

xmbrst

  • 13 Posts
  • 3 Reply Likes
Voting up of all-GC designs is indeed a problem. I see a flip-side to the problem, too, though: given the consensus among experienced players that lots of GCs are bad, how could anyone get a design synthesized that tested whether there were better or worse ways to use GCs?

In other words, lets say I come up with some clever way to pattern a design with lots of GCs that avoids the sticky glue problem. How could I ever get this design synthesized in a world where to earn the privilege of voting players have to agree that GCs are bad in some absolute sense?
Photo of Fomeister

Fomeister

  • 12 Posts
  • 4 Reply Likes
The problem is this: The "consensus among experienced players that lots of GC are bad" is absolutely positively NOT TRUE.

Yes, in certain incidents of the labs certain designs did not synthesize properly with "a lot" of GC's.

Consensus is not supposed to be the deciding factor. The lab _results_ are the deciding factor.

Perhaps the most telling statistic you should consider it this one.

100% of the people in this project, including the lab, have no idea what works.

So if you think that precluding design on the basis of "consensus" is the way to go, Make a Group. Join the Group. Reach a consensus, and submit your designs.

IF, your observations are true, 100% always, than the game mechanics are easily modified to prevent too many GC bonds.

But they are not, and for good reason.
Photo of Fomeister

Fomeister

  • 12 Posts
  • 4 Reply Likes
Just one more comment on this...

Does anyone look at the winning design at all?

Because if you did, you can see that the BEST lab results came with the 2 designs with the highest GC total. Excluding the all GC design of course.

So can you explain, intelligently, what is your problem with GC's?
Photo of Matt Baumgartner [mpb21]

Matt Baumgartner [mpb21], Alum

  • 128 Posts
  • 33 Reply Likes
In general the problem with having a RNA with all GC's is that it lacks sequence specificity. Meaning that it is difficult to make patterns of strands that will only bind to their target strand, and not mispair with other strands. If you add in other forms of nucleotides (i.e. AU pairs) then you can make strands that have higher specificity.

That being said, I do think that it is possible to make a properly folding RNA strand using (almost) all GC pairs. But you would need to make extra sure to have high specificity.
Photo of aculady

aculady

  • 10 Posts
  • 6 Reply Likes
If the rules about how RNA folds in the lab were simple, we'd already have figured them out and this project would be unnecessary.

I think it is very dangerous and, frankly, premature to start putting out any authoritative set of guidelines that encourages people not to explore the full range of possibilities available in the game.

Synthesizing RNA that does not happen to fold properly is NOT a waste of lab time. Seeing how and why things go wrong is one of the most powerful ways we have of deriving rules about how and why we can make things go right, and the evidence of one or two designs is not enough data to say that we have definitively proven anything, because each design includes many different aspects that might influence folding. Knowing if we are seeing the results of bond repetition, free energy, sequencing issues, or other factors requires seeing the results of multiple iterations of similar designs where, to the largest degree possible, we are assessing the impact of changing a single aspect of the design at any one time.