Energy in neck area

• Article
• Updated 7 years ago
Some time ago when we did not have working neck in the branches design lab puzzles, I predicted on the basis of my even energy theory, that energy content in the neckarea needed to go down.

It turns out I was right – on average that is. The later working necks were around 2 energy points lower than the non-working ones. Here is my datasheet with numbers for both working and non-working necks.

Collective energy in necks for highscoring Branches designs *

My prediction holds, despite I made it on the wrong presumption that energy would be evenly distributed in the neck area too. It turned out it wasn't. The collective energy in the necks were ranging between -8,4 to -18,8. The lowest I call low energy neck, the highest I call high energy neck. All necks ranging between 12 and 16 I call middle energy neck.

As I mentioned in my getsat post What is so special about the neck area?, the neck area is the only place in a rna design where it is allowed to go cub scout or christmas tree, if following a certain pattern.

But there might also be a pattern when it comes to collective design energy and collective energy of the neck. We might not yet have data enough to paint the whole picture. So here is a quick outline of what these tendencies might be:

It seems a middle energy design, will rather have a high energy neck or a low energy neck, than a middle energy neck.

And from what I have seen so far, I think there will be a pattern of a low energy neck together with a low energy designthere seem to be a pattern of a low energy neck

Also a low energy neck will often come with a lowered energy value in the multiloop and/or hook. That is, the GC-pair in the will turn opposite and lower negative energy in the multiloop.

See Aldo's Cyborg 2 (100%)

This is a low energy neck (-8,4) in a fairly low energy design (-23). And the GC-pair in the multiloop/neck turns opposite, lowering the more usual multiloop value -1,1 to -0,8. Also the energy in the hook is lower than usual, a more common value is -1,9.

*Note, two of the branches designs (Mat branches – small mod of my 95% and Bonsai sapling by Paul Denkmann) have not yet been synthesised and are therefore not in the datasheet. I don't think two designs are enough to change the overall tendency.
• 2223 Posts

Posted 8 years ago

• 549 Posts
I have another look at energy in the neck, related to the energy bonus gained by forming the closing pair. It is based on looking at the energy model rather than the lab results. It might be interesting to see if the prediction that it makes (that G..C and slightly less so, C..G are the strongest neck closing pairs) is backed up by lab data.

• 2223 Posts
Hi JL!

Here is the lab data part of your energy model data. Turns out that the energy model is overall correct at this area.

Energy in neck aligned with energy spots in the design
• 549 Posts
I'm curious whether having a lower energy neck might be helpful in that it allows the neck to form later rather than sooner. If a shape forms from the hairpins back to the neck then there are fewer large loops formed early. Large loops have high free energy, so there is an energy barrier to their forming, and if they happen to miss-form, there is a barrier to their undoing (at least according to the energy model).

So having a lower energy neck will likely mean fewer strong stack quads in the neck, which likely lessens the chance of mismatch or early match and allows the shape to form from the end-loops back towards the neck.

In contrast, having stronger hairpins could allow formation of the shape starting with the end-detail off of an open loop without the energy barrier required to form a large loop and then sub-divide it into stacks and loops.

So the theory would be that it could be helpful to have stronger hairpins and a lower energy neck -- but how to test that?
• 2223 Posts
Hi JL!

There are some interesting questions there. The other day in chat Dwar gave an explanation on how RNA folds up, that I think was intresting. I think you will find it interesting as well.

Here is a cut from the chat:

Dwar: i know for a fact the "solutions" i come up with wouldnt actually fold this way [8:23 AM]
Matthew Leslie: how do you know [8:24 AM]
Dwar: it exploits the computer's way of figuring it out, there are weird factors in play in real life. like sequences that are close to each other will fold up first [8:23 AM]
Dwar: which this doesnt account for [8:24 AM]
Dwar: what we find might have the most free energy, but that doesnt mean it will fold that way [8:24 AM]
Matthew Leslie: hmm what you say makes sense but how do you know this game doesn't account for that [8:24 AM]
Dwar: because, sometimes the RNA will misfold [8:24 AM]
Dwar: im pretty confident it doesnt [8:24 AM]
Dwar: from working with other software for the same thing [8:24 AM]
Matthew Leslie: i wonder why they'd omit something so obvious [8:25 AM]
Dwar: i think it's just too hard to do [8:25 AM]
Dwar: i dont think it's that they overlooked it [8:25 AM]
Dwar: like in nature RNA will fold 5' to 3' [8:25 AM]
Dwar: or from 1 - N [8:25 AM]
Matthew Leslie: what does that mean [8:26 AM]
Dwar: because that's how it's synthesized in the cell [8:25 AM]
Dwar: so the beginning gets made first, so it will start folding up before it's done being made [8:26 AM]
Matthew Leslie: oh totally [8:27 AM]
Matthew Leslie: that makes sense [8:27 AM]
Matthew Leslie: but then one side would start folding first not in the middle [8:27 AM]
Dwar: exactly [8:27 AM]
Dwar: which it gets "trapped" [8:27 AM]
Dwar: so it doesnt fold in the most energetically favorable position [8:27 AM]
Matthew Leslie: because the whole strand isn't available at the same time you say? [8:28 AM]
Dwar: exactly [8:27 AM]
Dwar: nice [8:27 AM]
Matthew Leslie: i'm an EE so all this stuff is new to me [8:28 AM]
Matthew Leslie: why don't they 'brace' it until the whole thing is available [8:29 AM]
Dwar: when they test it ( im guessing because i dont know the exact experiment they run) i think they heat it up and cool it down [8:28 AM]
Matthew Leslie: is that not possible [8:29 AM]
Dwar: so the whole thing is still available [8:28 AM]
Dwar: which is a little different than in nature [8:28 AM]

So I think RNA behaves a bit different in lab, from how it does in the cell. That is what I read from this.

And your question on if low energy necks do better than middle and high energy ones. I think a bit of statistics will help show

I colored my spreadsheet from an earlier forum post made in an attempt to answer that question.
• 549 Posts
It makes sense that in nature RNA starts to fold as it is created, but even so, unless there are extremely strong bonds it can unfold and refold. Still, fold as you go should favor stronger hairpins and (somewhat) weaker necks, or at least not be disrupted by such.
• 6 Posts
Hi, all.
I've been around here less than a week a I find this all so fascinating.
Congratulations to all.

As I read jandersonlee's comment above about sequential folding I thought it would be interesting to follow a sequence folding from start to finish.
So I took Aldo's Cyborg 2 and did just that.

I had suspected that EteRNA least energy solution for each stage would be radically different but, in fact, the shape evolved quite smoothly.

To answer jandersonlee's question, in this case, it was the hairpin that formed first.
Here are the images of the jump that forms the second hairpin between stages 53 and 54.

• 6 Posts
... and I suppose that it's no surprise that it was Cytosine @54 that shook things up.

I know I should have made a note of the energy at each stage, but I do remember that there were no big numbers and / or changes as the sequence progressed.
I'm not sure if this "result" extends outside the "game", but I it would seem to support Eli's preference for smooth energy .
• 6 Posts
... and here is the rest of the branch falling smoothly into place at 59. (and no prizes for guessing what colour 59 is!)

• 2223 Posts
Hi byRo!

Welcome in here.

Just to make sure I understand. You filled out the actual lab sequence for Aldo's Cyborg 2 and filled in from nucleotide 1 and up, to see how the energy model in the game would react?
• 549 Posts
Fresh blood, fresh ideas! Thanks byRo for doing that and sharing the results!
• 549 Posts
I tried this idea of assembling an RNA sequence one nucleotide at a time on my tweaked mod 93 submission for Round 9 of the Simple RNA Switch Lab.

You can see the sequence in this Google Doc

Basically it starts as a straight line, folds into the bound form at one point, then into the unbound form as it progresses. No weird intermediate shapes. I hope that's a good sign!

And yes, Eli, I understand this is only what the EteRNA model predicts, not what happens in the wet lab, but it's what we have to work with!
• 549 Posts
Also in terms of necks, both the bound and unbound forms came together *before* the final nucleotide to close the neck was added. Thus it seems that, by the model at least, a strong neck closure is not necessary to form a shape. In fact, the final neck bond may sometimes be the last one to form. Whether or not this is typical, I don't know.
• 6 Posts
I found a quicker way of doing this today. (or, maybe, this was what you already did for your RNA)..

- Copy the RNA sequence to Notepad (or similar, but it must be monospaced);
- Write a line of dots underneath exactly the same length as the sequence;
- Open puzzle creator in EteRNA;
- Copy the dots to the shape data;
- Copy the sequence to the sequence data (duh!);
- Set to "Natural" mode (you should now see your RNA fully folded)
- Position the cursor at the end of the shape data dots and backspace....

(yes, it's back to front, but fun still the same)
• 549 Posts
Nice tip. Thanks.
• 6 Posts
As to the relevance of this, I believe that it depends on how you consider that RNA folding actually happens.

A) If you believe that folding happens sequentially from the start of the sequence to the end then this information is extremely relevant as it gives you an idea of the path through which the RNA arrives at the final shape (which it is not actually constrained to coincide with MFE).

B) If, on the other hand, you believe that RNA just happens to fall into the right place (MFE) all at once, then it doesn't matter at all where it might have been when only half the sequence was available.

It comes down to the difference between a film (A) or a photograph (B).

I'm A) !!

(There's probably a C) but I'll leave that to the experts)
• 549 Posts
There is a C: you can partially explore it at the barriers web server.

Basically most RNAs are constantly folding, unfolding and refolding. Each conformation (shape) has a given free energy and a probability of forming based on that free energy. In a given in vitro "soup" containing of billions of identical RNA strands, different strands may be folded into different shapes at any one time.

The RNA molecule will tend to prefer shapes with lower (more negative) free energy. However, Brownian motion of the molecules in the solution that contains it can sometimes bump it into a new configuration with a higher free energy state.

The shape changing tends to happen just one or two bonds at a time though, so to pass from one low-energy form to another it may have to pass through one or more (sometimes several) intermediate shapes with higher free energy. When there is an intermediate form that has a significantly higher energy than either of the lower energy shapes, that difference in energy acts as a "barrier" to the molecules transitioning from one shape to another.

So although the EteRNA game shows you "one" shape for any given RNA strand, there may be many possible shapes that can exist. The challenge of the lab is to find a sequence that makes the desired shape so much more likely that others shapes are rare, and with few high-energy barriers between different low-energy shapes (otherwise some molecules could get "trapped" into avoiding the desired conformation).
• 6 Posts
OK got it.

But that's something that bugs me. If many of the RNA strands in the "soup" are folded into different (wrong) shapes then how do they do their job properly?
• 549 Posts
If they are folded wrong, they don't. But evolution tends to select strands that do a good job a sufficient amount of the time.

The stand-in for evolution in the labs is people who take high scoring designs from an earlier round and try to make small changes to improve their scoring. Like any mutation, sometimes they do, and sometimes they don't. But as we learn more about how RNA performs in the labs we can (a) do a better job with the first few rounds including (b) by reusing forms from earlier labs that worked and (c) making more intelligent modifications.
• 2223 Posts
Hi ByRo and Jandersonlee!

Thanks for bringing up this neck topic up again. It made me take a look at my old colored spreadsheet that I left months ago. I knew there was some patterns, I was just not sure how to bring it forth. Also it was not an overall pattern.

This time I cut out the two columns that looked like they correlated most. So far it looks like a low energy neck often, but not always, comes in connection with a low energy design. And a high energy neck often, but not always, comes in connection with a high energy design.

Almost half the winning designs with working necks have that high energy neck/high energy design pattern, while a little less than half have the low energy neck/low energy design. Or a bit more precise, of the 77 winning designs with working necks, 37 (48%) of them show the high energy neck/high energy design pattern, 29 (38%) show the low energy neck/low energy design pattern. Only 11 (14%) designs show a mixed pattern. Of them 6 (8%) have low neck energy/high energy design and 5 (6%) high neck energy/low energy design.

So far I can not say if low energy necks overall will be more successful compared to high energy ones, though I do suspect so. (At least for the longer stringed ones.) I don't have enough data yet to prove it. What we have of low and high energy necks still to some extent mirrors what players have a tendency to use. I just noticed that we seem to get a higher success rate for necks after players started copying Nupacks successful low energy necks.

I'm not sure I will make a market strategy banning designs that does not have this pattern low energy neck/low energy design and high energy neck/high energy design, as that will rule out perfectly good designs that do not follow it. A low energy neck can be perfectly fine in a high energy design and opposite. If we could just predict when we need what combo, a lot would be gained. And I think we might one day be able to.

There seems to be a pattern, but it is not totally permeating. Perhaps I will instead try reward designs that have this so far strongest pattern.

And as Neil Mussett mentioned something about standard mean deviation in the forum lately, it rang a bell. So I have taken the highest and lowest neck energy for a lab and found the mean, and distributed the colors accordingly to that. Same with total energy. It actually made the pattern a bit clearer, than my previous estimated color distribution.