For theory behind this see my forum post Energy in neck area

I would like a strategy that says:

Give + 1 if both total energy of a neck and total energy of a design is low

Give + 1 if both total energy of a neck and total energy of a design is high

Count this for a low energy neck: If it has an energy lower than mean of the total energy of necks in that lab.

Count a design for having low energy if it has a lower total energy than the mean of energy for designs in the lab.

1 Like

Hmm. I get the idea of balanced energy in the neck and elsewhere, but can you make it self-referential rather than relying on other designs? Like the average energy for stack cells in the neck in relation to the average energy of stack cells elsewhere in the design?

Let’s suppose that an optimal neck should have an average stack cell strength of K times the average stack cell strength elsewhere in the design. We would penalize a design P points for every 1 SD of average stack cell strength elsewhere that it differed from this. (Data mining of old designs might help us find values of K and P.)

In Aldo’s Inspiration by starryjess, the average stack cell strength in the neck is -1.97 while the average strength for stack cells elsewhere is -2.23 with a sample SD of 0.499. For the sake of example, let’s assume that K is 0.92 and P is 25.

abs(AvgNeck-AvgElsewhere*K)/SDElsewhere

= abs(-1.97 - (-2.23*0.92))/0.499

= 0.086/0.499

= 0.172

So the design neck is within 0.172 SD of the expected strength. The score would be:

100 - 0.172*P = 100 - 0.172*25 = 100-4.31 = 95.69

Of course, I picked K and P so that the score would come out about right, but you get the idea I hope.

Note, I’m not sure whether the penalty should be based on the average stack cell strength elsewhere or the standard deviation. The more I look at it I think it should be the average. Otherwise designs with more uniform stack cell strength are penalized much harder for variations in neck strength than designs with varied stack cell strength.

Hi JL!

At first I didn’t understood what you brought up. But I think I do see a problem with my strategy. I am assuming I beforehand have all the data from a full lab of the same shape, where know what is a low energy neck/high energy neck and low energy design/high energy design. Those numbers the bot will first have after a full lab. So I do see a pattern, but I am not sure how to make use of it for a strategy.

But if I can get some sort of mean value for a high/low energy design, as you suggest, I might get somewhere. Only thing I worries a bit about is that short necks behaves a bit different to long necks. In other words, the short might have a different energy density compared to the long ones. The short necks seem to be able to handle more GC-content than the long stringed ones. But in my data I have still not worked out if the pattern I found is dependent on length of neck too. I have considered looking into it. And I guess that this flawed strategy of mine, is a quite good reason to do it. If I could rule out that energy in necks needs a different energy balance depending on it being long or short, I should be able to use your mean energy of neck idea to say something about what makes a neck and design energy rich or not.

Dear Eli,

Your strategy has been added to our implementation queue with task id 140. You can check the schedule of the implementation here.

Thanks for sharing your idea!

EteRNA team