Something loopy – energy in multiloops (Branches design)

  • 1
  • Article
  • Updated 8 years ago
(Loops from which more than one stack stems are called multiloops, source Jeehyung)

I looked at the values in all the synthesised designs in the new advanced mode displaying free energies.

Just two of the bots manage to keep one of their three multiloops together. (Vienna 5, 81% og Nupack 1, 80%) Players manage one or two. Mat, Chaen, Ding and Starry are the only players with two intact inner loops.

I think there is something really important going on in the multiloops when it comes to energy. It might be one of the main causes why this multiloop design are so hard to keep together.

In general, when I compare energy levels in multiloops in human and bot designs, bots use a wider range from +4,3 to -0,7 using more and higher positive numbers than players. Energy levels in player designs range from +0,7 to -1,4.

Players have the highest negative number in the biggest multiloop (-1,4) and bots have the highest positive number in the biggest multiloop. (+4,3)

I haven't figured the secret of the multiloops out, but I see one clear tendency. I was able predict in which puzzles I was sure all three multiloops would break. All the robot designs with 3 positive values in the multiloops broke apart. (Will this be a generel tendency?)

I think the energy values in the multiloops of the robots, are a bit too low and that they need to get higher negative numbers.

Of course we need to see more synthesised designs. I look forward to see where this is going.

Two multiloops can stay together, with two numbers with low energy(+numbers). (Mat Branches V1, 85%)

What makes lower energy in multiloops (positive numbers) or what not to do:

Using AU-pairs in the multiloop as the closing of a string
Not having GC-pairs turn in the same direction
Placing lots of nucleotides in the ring of the multiloop (a few green or red at the right side of the GC-pairs in the loops, as Ding mentioned, will probably be okay)

Tell me what you think about this, have some of you discovered other things concerning multiloops?
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes

Posted 8 years ago

  • 1
Photo of Jeehyung Lee

Jeehyung Lee, Alum

  • 708 Posts
  • 94 Reply Likes
I like the idea of looking things in energy perspective -

Have you looked at a design by NUPACK bots? one of them had a successful neck, but the neck's free energy total wasn't that low (few AUs in the stack)? Very confusing result there.
Photo of Eli Fisker

Eli Fisker

  • 2223 Posts
  • 485 Reply Likes
Did you mean Nupack bot design 6?

I think I understand why you are confused.

I get a different energy value, when I ad up the numbers in the neck in the new advanced eterna mode, than in the genebee tool. I just checked with values from eterna, as for some designs, if genebee thinks they break up, it don't show the collected energy for each arm, as some of them melt together or combine with loops. This happens in the designs I mark *.

Collected energy in working robotnecks

Robot Eterna (Genebee)
Nupack bot 1(80%) -8,5 (-8,4)
Nupack bot 3 (69%) -14,9 (-13,8)
Nupack bot 6 (67%) -16,4 (-14,5)
Vienna 5 (69%) -11,4 (-7,2 *)

So I thought the negative energy level in the robot necks were lower than it were. Though clearly not perfect, it was the genebee tool that made me realise that there might be something hidden in an even distribution of energy. In this case Genebee show smaller negative numbers for bots and higher negative numbers for human designs, compared to the correct eterna energy score. So this energy tendency I noticed between human and bot design, is not as strong as I originally thought.

We only have partial working necks from players so far in the branches design, though one almost made it.

Energy level in necks from player puzzles (though not working)

Player, design Eterna (Genebee)
Mats branches v1 (90%) -15,4 (-16)
Mats Branches v1.1 -15,4 (-16)
Starryjess, Y oh Y (84%) -15,2 (-15,6)
Penguian, Toad hall (83%) -15,5 (-16,1)
Chaen 202- (83%) -13,3 (Partly working) (-13,1)
Jonathan Hall (81) -12,9 (working save one basepair) (-5,3 *)
Ding, Ding's branches (80%) -13,5 (-14,4)
Aldo, Modolus (79%) -13,2 (Partly working) (-14,1)
JerryP70, JP-11-4-18 (79%) -13,3 (Partly working) (-12,8)
Eli Fisker, Branches (76%) -15,5 (Partly working) (-16,1)
Starryjess, Starrys branches II (76%) -15,3 (Partly working) (-15,6)

I think there could still be somthing in the energy difference in necks between players and bots. Because in the best human designs with nonworking necks the negative energy is higher than for half of the bots with working ones. And some of the partly working player designs have smaller negative numbers, than the nonworking ones.

I was surprised when one of the robots made a working neck with only 2 GC pairs and the rest AU pairs. (Nupack bot design 1) I think this is too many AU pairs, but it made me realise that maybe we put in too many GC-pairs, just to be sure that the neck stick together. Though I'm not as certain as before that this theory will be proven right.