Messy pdb structure

[bas-rustenburg]

I'm currently debugging my attempted constant-pH simulations, many of which crashed for unknown reasons. I've found out however that at least one of them might be the result of a bad pdb structure.

This structure seems to have a messed up segment approximately between residue numbers 186-201: https://github.com/choderalab/mcce-charges/blob/master/pdbs/Crizotinib-ALK/2XP2_fixed_ph7.4.pdb

I'll respond to this thread if I find any other dubious structures.

edit:

The pdb structure before "fixing" looks fine.

[bas-rustenburg]

This is most likely the result of pdbfixer modeling in a missing loop.

[bas-rustenburg]

Afatinib EGFR, 4G5P_fixed_ph7.4.pdb has a weird loop as well. (Resi 295-315 in that file)

[bas-rustenburg]

There is even a leucine with the wrong protonation state,,,

[mrgunner]

Will it run with something that is simple? eg BPTI (very small) or Lysozyme (the standard pk benchmark protein) (these are also good in that we know the pKas.

On Oct 6, 2016, at 11:47 AM, Bas Rustenburg notifications@github.com wrote:

There is even a leucine with the wrong protonation state,,, https://cloud.githubusercontent.com/assets/3977731/19159473/a8a99b8e-8bba-11e6-95fc-5b6b0dd164e2.png — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/choderalab/mcce-charges/issues/47#issuecomment-252003920, or mute the thread https://github.com/notifications/unsubscribe-auth/AJtPXjmdYvsPuxct2G-PdPsODhZWHrkHks5qxRgXgaJpZM4KQAQR.

[jchodera]

Are there missing loops in BPTI or lysozyme crystal structures? I think these are small stable proteins where the residues are generally well-resolved.

[mrgunner]

There are about 1700 structures. The pK calc have often started with 4LZT. I think it’s complete. 129 residues

Just for fun you can look at the 2nd page of this excel sheet where we pulled out structures crystalized at different pKas or salt.

On Oct 6, 2016, at 3:05 PM, John Chodera notifications@github.com wrote:

Are there missing loops in BPTI or lysozyme crystal structures? I think these are small stable proteins where the residues are generally well-resolved.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/choderalab/mcce-charges/issues/47#issuecomment-252058679, or mute the thread https://github.com/notifications/unsubscribe-auth/AJtPXgU4GQ-qht-w2mJdhrOfMmRmtmi7ks5qxUZ6gaJpZM4KQAQR.

[steven-albanese]

@bas-rustenburg: I've checked in renumbered PDBs in PDBFinder that were all created from the same fixer.topology and fixer.positions objects (i.e., there shouldn't be any more differences from loops being built in differently between apo, apo-nowater and fixed structures). There's definitely still an issue with loops being built in poorly, so I'm looking at PyRosetta to see if that might be useful to build the loops in more reasonably

[bas-rustenburg]

I've tried simulating 4LZT using the constph code, as @mrgunner suggested. I ran it for 60 ns and encountered no issues. That makes me more confident about the constant-pH code. Our issues are probably related to the starting structures.

Discussed with @jchodera and @gregoryross today. We're going to try ensembler loopmodel on the kinase structures since pdbfixer ended up producing problematic structures. I'll have a chat with @steven-albanese on how we'll handle it. If we can validate that those structures look more realistic, we should also redo the MCCE calculations using those structures instead. It would be best to not rely on the results that were using pdbfixer structures.

[mrgunner]

that’s great. We can give you the experimental pka’s for this if you want (salah - ask divya for them)

On Oct 18, 2016, at 3:46 PM, Bas Rustenburg notifications@github.com wrote:

I've tried simulating 4LZT using the constph code, as @mrgunner https://github.com/mrgunner suggested. I ran it for 60 ns and encountered no issues. That makes me more confident about the constant-pH code. Our issues are probably related to the starting structures.

Discussed with @jchodera https://github.com/jchodera and @gregoryross https://github.com/gregoryross today. We're going to try ensembler loopmodel on the kinase structures since pdbfixer ended up producing problematic structures. I'll have a chat with @steven-albanese https://github.com/steven-albanese on how we'll handle it. If we can validate that those structures look more realistic, we should also redo the MCCE calculations using those structures instead. It would be best to not rely on the results that were using pdbfixer structures.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/choderalab/mcce-charges/issues/47#issuecomment-254618130, or mute the thread https://github.com/notifications/unsubscribe-auth/AJtPXuQaChZyBL9HF1oB7B64sWq3Rom6ks5q1SIUgaJpZM4KQAQR.