(back to Conformational 4)
You may wonder why I'm starting this chapter with this video. Well, the MinutePhysics channel is quite awesome, for one. But it also displays graphically what I'd like to talk about here: energy barriers (the mountain pic) and local minimas (the valleys).
Back to JR's design. One of the simple explanation for the experimental result would be that our software model is flawed. It is after all a prediction that is known to be moderately accurate. Possibly, the FMN-binding shape could actually be more stable than the calculated MFE .
But, do we absolutely need to question the simulation in this case? Aren't there mechanisms that would allow that the experiment tells us "FMN-binding shape", while the predicted MFE is also energetically more favorable? As graphically pictured in the video above, it could be, in the form of valleys and mountains in the energy landscape. Could it be the case in this design?
There's nothing really new about what follows on this page. The tool presented already existed back when EteRNA just started, and a knowledgeable player, alan.robot, had written a GoogleDoc about the topic (and more). I can't help feeling some regrets that he no longer participates in the game... If it hadn't been for another player, Eli Fisker, this contribution would probably be completely forgotten. I'd like here to thank both of them.
Sadly, we have to shorten the sequence, because of the 100 nucleotides limit, but for this design, we should still get rather accurate results (assuming the model is correct)
Erm... be patient. Or go get a coffee...
The results are apparently semi-permanent, they stay for at least days on the server, and this run can be found here:
The first thing to be found on the page is this tree:
The vertical axis is not too hard to figure out, it's the free energy. But what are those numbers at the leaf-points? They are identifiers for the secondary structures, and the complete list of suboptimals can be found here:
Unsurprisingly, the calculated MFE is the bottom-most leaf in the tree, denoted "structure number 1" . The structure number 2 is a close cousin of the MFE, and the third spot on the list is occupied by the FMN-binding shape .
Now to the funky stuff. I guess we're curious to know how the FMN-binding shape can evolve into the MFE. So let's just ask :)
And we get:
This piece of the page is really very cool, in my opinion. Play around with the controls, you'll see.
Supposedly, the calculated path is the "best" one between the structures. Assuming this is correct, it would seem (marked area), that there is a possibly serious difficulty...
Ok, difficult. But, how difficult can it be?
Well, the page has another nifty tool. It can tell you in what shapes a specific state will evolve over time.
So let's try a few simulations. We take a sample of the RNA sequences composed exclusively of the FMN-binding shape (structure 3).
If we ask what shapes represented 10% of the total at any point in time over a span of 1 million seconds...
we notice that the MFE (structure 1) is not there...
Hmmm, what about 5%?
Come on, at least 1%, right?
After 1 million seconds (calculate how many days that makes and wonder if the guys in the lab, or even us, are willing to wait for that long), no MFE appears from a pool of FMN-binding shapes. None.
Let's try to wait longer.
Ahaa... here it is. After 200 million seconds (a bit more than 6 years), the MFE has finally become the dominant species in the solution. And at 1 billion seconds (more than 30 years), it has still not reached equilibrium...
The conclusion of this simulation is that we don't need to question the Turner model, at least, not for this design. If the RNA sequences reached preferentially the FMN-binding state while folding, then it is very normal that they stayed in that state and didn't evolve into the MFE structure, because of an insurmountable energy barrier.
Now, how on Earth did they get in the FMN-binding state in the first place??