Eterna3D Ribosome and beyond

We got our first look today at the new 3D model being developed for in-game visualization. In case you missed the town hall, here are two images the developer Pragya shared with me.


I’m wondering if tertiary bonds within the rRNA are shown on the 16S model? (Not the contacts with proteins, the 16S-16S contacts.) And I’m wondering if there is any way to use our existing motif/protein booster to light up motifs on the 3D model. Since Pragya is doing this amazing programming, maybe she knows a way to do it that wouldn’t use much processing power. The design decisions being made likely are mindful of processing burden; don’t want every new mutation to take 5-10 seconds to fold.

First a massive thanks to Pragya! You rock! I think the new 3D interface is beautiful. Eterna flat RNA society joins the spherical age.

Unfortunately this thought first popped up after the townhall. I think I immediately got distracted by the fine M&M/smarties chocolate bases. Anyway here it comes.

I think there is one thing concrete that would benefit player understanding of how the bases fills in space.

The smarties are all the same size. But I would like to have the C and U be half the size of the G and A, because that is more realistic.

Chimera also have this for its colorful stick version

billede
Image from Omei’s Introduction to Chimera for EteRNA Players (page 11)

The bases are not of equal length. A and G are long bases and C and U are short.

A while back I wrote a story about it.

DigitalEmbrace asked me: Please confirm for me the second image is a pseudoknot.

This one from her post at the top:

Eli: Do you mean if it is a pseudoknot in 3D?

DigitalEmbrace: I’m 99% certain that is a pseudoknot in 2D (and 3D) but tell me now if I am mistaken.

Eli: I can find 4 tails
actually make that 5

I think it may be something about this special puzzle type that makes the structure break up in tails in 3D. There is probably some programming to be fixed for the two puzzle types to speak together between 2D and 3D. Nicely identified.

Your image highlights that there may be a problem.

And I have identified where the problem is

So the gap between the straight strand and the curvy single bases in the left image, is where the backbone breaks in 3D. (green ring) Basically between stem and curvy single bases

The one above is the one easiest identified

I identified 5’ in both puzzles and then followed the bases in 3D against the 2D.

RE: Tertiary bonds, the last we discussed this, we didn’t plan on adding them because as mutations are made to the 2D and reflected in the 3D, we don’t have any way to determine where the bonds should exist at that point - the 3D is purely “target mode”

RE: Base size, the different bases are already different sizes! It may be a bit harder to make out due to the rounded shapes being harder to compare at a glance, but if you look closely they are different.

RE: 3D structure of that particular puzzle, again we’re only using “target mode”, as defined by a PDB file (or something comparable) - any structural concerns are most likely due to the file Rhiju pointed to for that puzzle (or maybe the underlying RNA rendering library we use) rather than Eterna3D itself.

Using the 16S as an example, tertiary bonds exist in the live 16S the same way secondary bonds do. The tertiary bonds would behave on the 3D model the same as the secondary bonds, they wouldn’t change positions. When researchers and players have discussed the need for adding 3D to Eterna, the motivation has been to help players visualize the molecule in 3D space and to add tertiary pairing to more accurately reflect the true structure of the molecule.

If you mutate all bases to A though, even if the backbone stayed in the same position, surely some of those bonds would not exist, or new bonds would form (as the atoms available to form bonds would be different)?

Looks very interesting. Thanks for all the hard work.
Building the framework first is always a good idea without biting off too much.
If I understand correctly, for now the 3D image will always be “target mode” and not an 3D image of “natural mode”.

Later, what’s stopping you guys programming so that changes in “natural mde” are mapped to a PDB file format that could be submitted to a Eterna3D modeler.

Thanks

Basically, the issue is that there isn’t currently a “3D folding engine” that can predict a meaningful 3D structure in a reasonable amount of time. It’s something we’d like to integrate, but it just doesn’t exist right now. Maybe sometime it will though!

Is dumping or a “save as” option that maps “natural mode” into a PDB file possible. That file type seems to be accepted on many internet 3D rendering sites.

Just curious (LOL).
Thanks

I believe a PDB actually defines the 3D structure - which is exactly what we don’t know. :slight_smile: Correct me if I’m wrong here

I am going to do some research before I respond so I know what I am talking about. A few days probably.

Ok. You are going to have to put up with me because I am going to pick your brain and you may have to refer to the researchers what is missing.
Please read this article: RNApdbee—a webserver to derive secondary structures from pdb files of knotted and unknotted RNAs
see figure 3 in particular and tell me ( as a programmer) what is missing if we go backwards.
My guess is something “big” is missing. You have to tell me exactly what.
And then the question becomes can we “soft code” what is missing so Eterna players can fiddle with it. It’s going to be wrong but so what. Like a contact table? A lost cause for real researchers (it would be a time suck) but no real loss for a Eterna participant since the cost of a wrong map by a player would be pretty cheap in Eterna3D. I would see it as a first step in “collecting” whatever data is missing ( way down the road) and step by step correcting the original data. I would prefer your answer to address this infamous contact table I have in my head.

Looking at this file it doesn’t look promising.
https://www.umass.edu/microbio/rasmol/1fdl.txt

Some of the things you’re missing include hbond locations, nucleotide orientation, and even the precise way the backbone folds in 3D. Giving users a blank slate and nothing else doesn’t seem particularly viable for me, and having “bogus” information in the 3D seems mostly liable to lead to bad conclusions getting drawn, to me.

Consider that there’s an entire additional citizen science game dealing with predicting 3D structure - Foldit :wink:

There’s been some discussion of integrating rna_thread, but IIRC the speed/accuracy issue is still present there. I wouldn’t be surprised if at some point we see something integrated even if it’s only usable as a rare cross-check, but that’d probably need a far amount of additional R&D.

Yep, I agree with you. My direction is not worth the trouble.
Thanks for the discussion.

I think a different way of looking at the subject would to look at it like a Metro Rail Map. They do not reflect geography at all- they sacrifice accuracy in favor of being visually appealing. Its made to help the people who use the map navigate.

Play conclusions can only ever impact the game as the dev allow them too. Players are bound by the rules of the game universe, so even if the visual might lead to a wrong conclusion, but not allowing them to do anything with that bad conclusion, they will immediately know it is bad.

Remember its a game, and one the awesome things about games is that they have been making things 3d out of 2D visuals in some very creative ways. You could always just add a tilt button and make the puzzles look something like this if the goal is to convey overlap. :smiley:

Right - except we don’t really have such guardrails in the lab. Including this extra 3D information would be to allow players to make inferences based on things we currently don’t/don’t know how to encode in the game rules.

I do understand the rationale here, and I’ve been thinking further about it - I’ll have to discuss with the scientific staff about what is actually viable to encode.

My vision is that the tertiary bonds would be indicated by a solid line and static. They would not be interactive like the 2D bonds (assuming what you are saying about the 2D bonds is correct. I didn’t realize the 2D bonds would change in the 3D model.) I clipped an image of the 5S from a 2013 paper as an illustration of typical 3D bonds. Unfortunately, the 5S doesn’t have bonds connecting the helices, which is the aspect that becomes much more clear in a 3D model as we can see in Praga’s image above.

1 Like