There are four major issues associated with the recently released puzzle 445:
1: Malformed Asparagine– Asp 37 has the carboxyl end of the sidechain pointed at an extremely small angle to the base end. the amine end of the sidechain is affected, though by a lot less. See an image here:
http://fold.it/portal/files/chatimg/irc_99169_1312771984.png (credits to Enzyme for finding this)
2: This problem has to do with the amino acids 111 and 112. The problem is that the segment has a perpetually low backbone score, which is not solved by a rebuild of those and adjoining segments. In a related issue, the local wiggle function behaves erratically on those segments. Local wiggle moves the segments drastically, causing a dramatic score drop. This score drop is much greater than those sometimes seen on other puzzles. (credits auntdeen and myself)
3: The puzzle also has a very erratic opening. Shaking and wiggling the protein sometimes results in scores below zero, yet sometimes above 10000. Other results include 9680, 9740, and 9853. (auntdeen)
4: Finally, the puzzle also has threading problems. When threading model B onto model A, the protein has a extremely low score, and when shaken and wiggled out, extremely low backbone scores all over the protein. an image of this phenomenon: http://fold.it/portal/files/chatimg/irc_92665_1312773923.png (infajmc)
So far as we know, the bugs only appear on model B for 445.
Due to these errors, we recommend that you fix and re-release the puzzle.
I found a workaround for Problems #1 and #2:
After working on Model B for a while, save your current structure and load it as a guide. Then, apply an alignment using the "extend" template and then re-thread the protein to the guide. This will fix both the Asn 37 side chain and the bad backbone bonding angles between Ala 111 and Ser 112.
mmm asp is usually blue. the one at seg #37 on models A & B are blue. Not sure what the pic is of as it is brown.
Model A is really not doing anything at all so I would disagee with the statement that so far only Model B has problems. However I find the labelling on the aligments a bit mystical so could be wrong. Fairly obvious tho, wiggle one goes a bit green, wiggle the other, it stays red :-)
If you freeze seg # 109 & #114 then right click to wiggle just that section (at normal 1.0 CI), it will move around like a snake, scoring the entire puzzle in & out of the minuses. Current score at 10463, this will take it down as far as -6418. Fascinating to watch - but broken.
Spmm— the brown Asp is just a view setting applied to the protein; mine is otherwise normal, except for the error
Regarding #2: Those backbones are problematic but they can be tidied up with rebuilds. If you rebuild from #111 to the end (#121) you'll find that the backbone can move a lot more, compared to just rebuilding #111, #112 and their adjacent segments.
I admit I did some other funky things to make it work (such as turning those segments into sheets during the rebuild and attaching a band to encourage them to straighten out) but it is possible to improve the backbone at #111-#112.
Admittedly, this is probably the least serious of the problems that Thunk has documented.
Thank you for reporting this, I'm looking into what went wrong right now.
My apologies for this problem, these puzzles have been reposted and you should be able to load in your previous solutions.
It appears that the CLOSED first iteration of 445 has been awarded points.
Was this intentional?
Should those be wiped out since the reincarnated 445 is now active?
point were given , confirmed (in chat) that it is intended. not as many point as a full puzzle were handed
