Weird helix scores too well

Started by auntdeen

auntdeen Lv 1

There still seems to be something wrong with scoring.

Puzzle # 858 should be a basic refinement puzzle. However, this solution has horribly misshapen helices, but is high scoring - I ran a Quaking Rebuild on it which gained nicely but produced this mess.

We have all seen helices bend, curl funny, etc. But the wide looping at the top of the one on the left is just plain wrong.

This was in Main, Mac OS 10.7

AsDawnBreaks Lv 1

Yep, I've seen one that was a very nice shade of green that was bent at a 90 degree angle (maybe more, can't remember). I think that may have to do with the fact scoring is not based on SS… Though maybe it should be?

BootsMcGraw Lv 1

You might want to check the sidechains that make up the "weird" part. If they are not sidechains that are commonly found in a helix, perhaps that odd area might actually be a loop.

I always take given secondary structures with a grain of salt. They are only predictions, yes?

auntdeen Lv 1

Normally I would have done that… but this is the revisited Mouse Puzzle and usually the structures given in an old puzzle are mostly correct :)

bkoep Staff Lv 1

Thanks for bringing this up, auntdeen. It turns out there was a slight problem in the way the score function was set up for this puzzle (my fault, sorry). The problem only affects Puzzle 858.

drumpeter18yrs9yrs Lv 1

As an experiment, I'd love to see scoring based on SS. Maybe on devprev. Just to see the gains and losses and hoe much of a difference it would make.

AsDawnBreaks Lv 1

I would love a custom score weighting option in general, though the core wouldn't count. Also, maybe being able to have tools act differently depending on the score weighting being used (like CI does with wiggle, and other such tools).

bkoep Staff Lv 1

drumpeter, can you elaborate on your idea of "scoring based on SS"?

Do you mean you'd like to see something that considers the helix/strand/loop "designation" that has been assigned by the player? This would mean, for instance, that your score could change simply by selecting a segment and changing its SS designation without actually moving the protein.

Or do you mean something that would reward segments with regular helix/strand shape and penalize segments with irregular helix/strand shape (like the funny helix in auntdeen's screenshot above)? This type of scoring is already used by Foldit to some extent, but could possibly be augmented in a devprev puzzle.