Closed since almost 4 years ago
Advanced Overall Prediction Electron DensityNote: This puzzle replaces the original Puzzle 2155, which was closed early due to a bug in cutpoint behavior. Players may load solutions from the original puzzle.
That is not a typo, this Electron Density puzzle has 588 residues! This will require using the new Trim Tool!
You'll also want to use the Radius Slider in the Electron Density panel, to view only density within a ball of the given radius. Remember that you select a sphere of residues with CTRL+SHIFT+click and drag, as demonstrated on the recent Lab Report! (Just make sure to click and drag on the protein, and not on the background.)
As in puzzle 2155, you'll see a gaps between segments 293 and 294 and segments 494 and 495. In 2155b, the gaps just show as empty space. The light blue "uncut points" from the previous puzzle are gone.
The comments for 2155 show the alignment of the Foldit protein to PDB 5DJW. The gaps are residues (segments) that weren't found in the crystallography results.
Using this recipe: https://fold.it/portal/recipe/46929 (the test lasts 10 minutes).
Result for a i9 computer (copy=pate from Note 1)(on reset puzzle):
Score:-401136.962 in 143 loops. Result: 4.12 sec / 1st loop
Why is there no leaderboard visible in the puzzle?
What do the trim subscores mean?
I started with a score of negative four hundred thousand. I clipped a section of about forty segments that were wickedly red and wiggles them to a score of eleven thousand. When I rejoined the trimmed section, my score was… negative four hundred thousand.
What is the purpose of the subscore that was visible? It seems meaningless in this context. I suggest also showing the untrimmed score at the same time, so we can see that we're not wasting our time cutting these monsters into smaller shreds.
The ferrets burrowed this bug really deep, but it should be fixed now.
Thanks for your patience!
You're right, when the Trim tool is active, the top-line score does not have much significance. In fact, with the next software update, the score will be clearly struckthrough and marked invalid while the Trim tool is active.
As long as the Trim tool is active, the top-line score is a very misleading guide for "Am I improving my solution," because it does not include energies for the hidden residues. (This is where the performance gain comes from. If we wanted to show the untrimmed score at the same time, Foldit would necessarily become slower because of these energy calculations.)
Folding with the Trim tool will probably have to be self-guided. For that reason, the Trim tool is not recommended for beginner Foldit players, and it will probably not mix with recipes that focus exclusively on score.
However, you are not left completely on your own. You can still pay attention to per-residue subscores like Density or Backbone (visible by pressing TAB while the cursor is over a residue). Or, I can imagine Trimming down large solution to fine-tune a H-bond Network in the core, for example.
See also this comment on the blog with a hypothetical scenario where the Trim tool might help hand-folding.
This puzzle crashes when I run scripts (even when trimmed to exclude "missing" residues), and even crashes when I wiggle all.
I downloaded the most recent devprev version. Nothing changed. No, I don't have a log. I'll post it, next crash (which should be in the next twenty minutes).
Dear Foldit Overlords,
The next time you throw a Godzilla-sized puzzle at us, could we please have two weeks to work it, instead of the customary one?
When scripts do work, they take exponentially longer to run than on smaller puzzles. We just don't have enough time to run what we need to run on these beasties to do these puzzles any justice.
All I did was a "wiggle all", and almost instant crash. Here's the log:
(edited down to traceback info, with long lines broken…long lines mess up the puzzle page formatting)
1: RaiseException +88 bytes (no line)
2: _CxxThrowException +101 bytes (no line)
3: __scrt_throw_std_bad_alloc +28 bytes (no line)
4: operator new +36 bytes (no line)
5: ObjexxFCL::FArray2D::FArray2D +93 bytes (no line)
6: core::optimization::Minimizer::dfpmin_armijo +266 bytes (no line)
7: core::optimization::Minimizer::run +1328 bytes (no line)
8: core::optimization::AtomTreeMinimizer::run +1010 bytes (no line)
9: core::optimization::AtomTreeMinimizer::run +31 bytes (no line)
10: interactive::application::actions::cart::PoseLoopThreadActionCart::run_minimizer_on_work_pose +1303 bytes (no line)
11: interactive::application::actions::cart::ActionCartGlobalWiggle::thread_body_loop_unlocked +162 bytes (no line)
12: interactive::application::actions::cart::PoseLoopThreadActionCart::on_thread_run +5761 bytes (no line)
13: ??$?RVToolActionThreaded@shared@application@interactive@@$$VP80123@AEXXZ$0A@@?$checked_fxn_wrapper@P8ToolActionThreaded
@shared@application@interactive@@AEXXZ@detail@util@interactive@@QAEXV?$access_ptr@VToolActionThreaded@shared@application
@Interactive@@@23@@Z +193 bytes (no line)
14: boost::detail::function::void_function_obj_invoker0<boost::_bi::bind_t<boost::_bi::unspecified,interactive::util::
detail::checked_fxn_wrapper<void (__thiscall interactive::application::shared::ToolActionThreaded::*)
(void)>,boost::_bi::list1<boost::_bi::value<interactive::util::access_ptr > > >,
void>::invoke +28 bytes (no line)
15: boost::function0::operator() +120 bytes (no line)
16: interactive::util::Thread::thread_entry_func +69 bytes (no line)
17: interactive::util::Thread::win32_thread_startup_helper +14 bytes (no line)
18: BaseThreadInitThunk +18 bytes (no line)
19: RtlInitializeExceptionChain +99 bytes (no line)
20: RtlInitializeExceptionChain +54 bytes (no line)
Long lines can make these puzzles pages difficult to use.
That's especially important on 2155b, where the scoreboard ferrets are unreliable.
In a rare display of moderatorory powers, I removed everything but the traceback from Boots' post.
I broke some long lines in the traceback as well.
The #bugs-and-feedback channel on Discord is a better place for this kind of feedback.
With the Discrod app, you just drag and drop your log.txt into #bugs-and-feedback.
(Still haven't tried that using the web version.)
Alternately, you can of course use the feedback area. But only replies to feedbacks allow file attachments.
(Original post available on request….)
