Empowered voting - calling close past lab results

  • 3
  • Idea
  • Updated 6 years ago
Just summing up what I’m asking about here. What I am asking for is a tool that when I look at a design and want to decide to vote for it or not and synthesized designs that are close in sequence. I will like a tool to give me the nearest designs (+ or - loop differences). It will be useful both when designing and voting. I will like to be able to compare designs that are close, one by one, beside the design I want to decide about. I think I got inspired from Jnicol's spreadsheets of the lab designs. There one can call out close designs.

Here is the mixed set of ideas that came up, when I mentioned this in chat:

Eli Fisker: By the way, I got an idea a moment ago [0:35 AM]

lroppy: brainstorm or just small shower [0:35 AM]

lroppy: ? [0:35 AM]

Eli Fisker: I am trying to pull op mods of the design I was deciding about to vote on [0:35 AM]

Eli Fisker: And it hit me [0:35 AM]

starryjess: OUCH [0:35 AM]

starryjess: ;) [0:35 AM]

Eli Fisker: why don't we get the designs presented that are already scored and 4 or 5 nucleotides away from that mod [0:36 AM]

Eli Fisker: instead of sometimes having to try guess what it is a mod of [0:36 AM]

Eli Fisker: I mean people don't always state what they mod, the precise title or what they change [0:36 AM]

Eli Fisker: And I like to see what is changed so I can decide if I think it is a good idea [0:37 AM]

starryjess: that is a good idea [0:37 AM]

Eli Fisker: Idea finished [0:37 AM]

Eli Fisker: thx, starry [0:37 AM]

Omei: I take a design with a good score, and then sort the designs on the sequence field. Is that related to your thought? [0:37 AM]

lroppy: again, concur [0:37 AM]

Eli Fisker: oh, Omei, that is a fine idea [0:37 AM]

Eli Fisker: but it takes more labour [0:37 AM]

lroppy: y [0:37 AM]

Eli Fisker: than I think it would have to [0:37 AM]

Eli Fisker: What if we could set how close designs we wanted? [0:38 AM]

starryjess: how do you do that Omei? [0:38 AM]

Eli Fisker: So eg. 5 bases different or 8 [0:38 AM]

lroppy: need more selectivity than that, i suspect, but on right track [0:38 AM]

starryjess: k nevermind, I see [0:38 AM]

Eli Fisker: So we can set how many alternative designs we want to see for a round 2 design [0:38 AM]

Eli Fisker: Iroppy, agree [0:39 AM]

Omei: And while they were add it, the developers could also not count hairpin differences. [0:39 AM]

Omei: * at it :) [0:39 AM]

Eli Fisker: Ah, very fine idea [0:39 AM]

Eli Fisker: I like that [0:39 AM]

Eli Fisker: Because they tend to vary more and will disturb what I am really after [0:39 AM]

Eli Fisker: I want to see designs that are similar in the main body, but slightly different. And I want to see the scored ones [0:40 AM]

Omei: Actually, that would be pretty easy to do as a script -- not have to wait for the developers. [0:40 AM]

Eli Fisker: Instead of having to search for the original for all the mods [0:40 AM]

lroppy: ooh oooh [0:40 AM]

lroppy: a volunteer coder [0:40 AM]

Eli Fisker: Omei, cool, go on [0:40 AM]

Omei: :-) [0:40 AM]

lroppy: would spend coins for that [0:41 AM]

lroppy: see another use for coins [0:41 AM]

Eli Fisker: hehe, definitely worth it [0:41 AM]

Eli Fisker: I will like to have it integrated in the game, so I am still considering putting it up in the forum. But I will love if you will work on it, or know someone who can, Omei [0:41 AM]

Omei: What parameters would you want to gice it? A design ID and a maximum distance? [0:41 AM]

Eli Fisker: A maximum distance, with and with out counting loops [0:42 AM]

Omei: It's the kind of thing I can do, yes. [0:42 AM]

Eli Fisker: awesome [0:42 AM]

lroppy: and ability to lock certain bonds so thaey do not change [0:42 AM]

lroppy: version 2 [0:42 AM]

Eli Fisker: When it gets integrated I will like it to be a tool where I won't have to give it an id, but just can call the tool [0:43 AM]

Eli Fisker: when in the design [0:43 AM]

Omei: Something like that would happen on the developer's schedule. But I could certainly prototype it, to help determine what the most important features would be. [0:45 AM]

Eli Fisker: That would be really great [0:45 AM]
Photo of Eli Fisker

Eli Fisker

  • 2203 Posts
  • 475 Reply Likes

Posted 6 years ago

  • 3
Photo of jandersonlee

jandersonlee

  • 530 Posts
  • 111 Reply Likes
> Omei: I take a design with a good score, and then sort the designs on the sequence field. Is that related to your thought? [0:37 AM]

> Omei: And while they were add it, the developers could also not count hairpin differences. [0:39 AM]

As I mentioned in another thread, a modified Hamming Distance option would help with this.

1) Pick a reference design/sequence
2) Specify a region of interest (default might be all but the bar-code hairpin)
3) Optionally specify fixed bases that must not change
4) Click to create a results display with a custom modified Hamming Distance column, counting the number of bases that differ from the reference sequence.
5) Sort on that column and the most similar designs/mods rise to the top of the list.

Having the ability to download a .csv of the submissions/results of a lab via a static URL might help those of us with some coding skills to prototype this as say a node.js Javascript service.
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
Rather than Hamming, I would prefer the algorithm I presented in https://getsatisfaction.com/eternagam... some time ago, specially for switches.
Photo of jandersonlee

jandersonlee

  • 530 Posts
  • 111 Reply Likes
@nando I like the idea of the non-symmetric distances. More flexible would be for a 4x4 table of distances (which could allow alternative weightings), but the weights/scheme you suggest could be a good default.
Photo of Eli Fisker

Eli Fisker

  • 2203 Posts
  • 475 Reply Likes
Wanted to add that Omei has written the script for calling close neighbours to a design. Here is a small description of how to use it.
Photo of jandersonlee

jandersonlee

  • 530 Posts
  • 111 Reply Likes
I may be missing something, but *where is it*?
Photo of Eli Fisker

Eli Fisker

  • 2203 Posts
  • 475 Reply Likes
Should probably try make it stand our more next time. It is hard to see. Glad you found it.
Photo of jandersonlee

jandersonlee

  • 530 Posts
  • 111 Reply Likes
Added a new script (tweak on Omei's) that uses Nando's distance metric.

http://eterna.cmu.edu/web/script/2713...
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
Cool :)
I added a comment to the script page. See http://en.wikipedia.org/wiki/Lee_dist... for why AC/CA should be 4 in this context.