extra lines drawn in Line+H and Stick+H views after mutate, hydrogen in wrong position; crashes while mutating

Started by LociOiling

LociOiling Lv 1

In a design puzzle, using the "+H" views (Line+H, Line+polarH, Stick+H, Stick+polarH), one or more extra lines (or tubes) are drawn from a backbone atom to one of the sidechain atoms.

Also, the hydrogen on the nitrogen following the sidechain is in the wrong position, appearing on the same side of the backbone as the following oxygen atom. (Hard to make good sheets that way.)

This happens for all amino acids except glycine.

Switching to a different view option clears the extra lines. Even switching to a different "+H" view clears the lines. For example, if you switch from "stick+H", to "stick+polarH", the extra lines (tubes in this case) disappear. The lines do not reappear if you switch back to the original "+H" view.

Once the additional lines are cleared, they do not reappear for the same amino acid at the same segment. They do appear when mutating to a different amino acid at the same segment, and for mutating to the same amino acid at a different segment.

Clearing the extra lines causes the hydrogen atom on the following nitrogen to flip back to the correct side.

This problem is generally reproducible, but not 100%. During testing on using a Contest, mutating to valine produced two extra tubes with stick+H on at least two tests. When I retested with the current design puzzle, 1355, valine only had one extra tube in some cases, and in other cases looked normal except for the hydrogen atom position.

In testing this issue, the client crashed several times when mutating. This may be a separate issue. An unhandled exception appeared in log.txt, but no debug.txt was generated. In puzzle 1355, mutating to glycine seems to cause a crash somewhat reliably. I'll open a new feedback if needed after further testing, since it's not clear whether this is related to having a "+H" view.

Release: 20170210-55c03a4ae1-win_x86

LociOiling Lv 1

After further testing, it seems likely that the crashes are confined to the "+H" view options, so I won't open yet another feedback.

The procedure to recreate a crash is:

  • open puzzle 1355
  • use selection interface
  • set view options EnzDes, Stick+H, Show All sidechains
  • select an arbitrary segment
  • mutate to glycine via M -> G
  • the client crashes

(It also crashes the same way with the mutate wheel in the original interface.)

I seem to have crashed the client with other mutations, but glycine is 100% reliable, while the others are hit-or-miss.

The crashes do not occur with other view options, including "Line" and "Stick".

In the previous post (and images), I believe I had the protein rotated 180 degrees, so the incorrect hydrogen is actually on the nitrogen in the same segment. (A careful inspection of the images seems to bear this out.)

Also, the bad hydrogen is most noticeable for valine and serine. For other AAs, it seems to become the extra tube or line, becoming bonded to the sidechain.

This bug may be confined to the "+H" views, but there's always the possibility that the root cause is something that's causing other problems.

Susume Lv 1

I was able to reproduce this crash on mutating to glycine with view options EnzDes, Stick+H, Show All sidechains, both in the latest devprev and in the one two updates ago (20170210-55c03a4ae1-win_x86-devprev), so it is not a brand-new problem.

In the older update, the first time I mutated to glycine I got a very weird screen rendering of the glycine, but no crash. I then changed it to valine and serine (seeing Loci's weird extra branches for both). Next I mutated to glycine again and it crashed.

Attached: log file from latest update, screen shot of the weird glycine rendering from older update, log file from older update. In the 2nd log.txt I removed several MB of nulls in two places. You have to search for the unhandled exception in that one because two other clients were also running in that folder.

LociOiling Lv 1

I've also tried to reproduce this in the "test" contest. The spurious extra branches still happen there, but so far, no reproducible crashes.

LociOiling Lv 1

I'm not sure if this is related to the other "+H" issues, but on puzzle 1378b, the ligand at segment 30 has a "shy" hydrogen that's too close to another atom. It looks like it's bonded to a nitrogen, but it's seems to be trying to burrow in to a carbon. Perhaps it's just a data file error, hard to be sure.