Turns out that the error wasn't so hard to spot after all. When extracting a pdb with cpptraj, it has problems when the number exceeds the width of the column. Once we reach 99999 atoms, it numbers the next one 1 (so 100001) effectively skipping 100000. Same with residue numbers once we reach 9999, the next one being residue 10001. I have gone back and checked my older pdbs automatically generated with cpptraj and realised that this had been happening constantly when the residue count exceeds 9999. I guess it was never an issue because beyond residue 696 those are all waters anyway, and LEaP did not seem to care the "missing" residue 10000 (and 20000, and so on). This seems to become problematic when it is also "missing" an atom and CONECT records are involved.
I don't know if this was a known issue for cpptraj or if it's just a fluke on my part.
Pietro
-----Original Message-----
From: amber-bounces.ambermd.org [mailto:amber-bounces.ambermd.org] On Behalf Of Aronica, Pietro
Sent: 01 December 2014 22:00
To: AMBER Mailing List
Subject: Re: [AMBER] LEaP Error
After searching, I have determined that the source of the error is probably within cpptraj rather than LEaP.
The CONECT records at the end of the file look something like this:
CONECT 323 1404
CONECT 2195 3168
CONECT 3615 4612
CONECT 4286 5057
CONECT 4420 5083
CONECT 5328 6325
CONECT 5999 6770
CONECT 6133 6796
CONECT 7224 8305
CONECT 910410077
END
They are generated automatically by a script of mine that calculates CYX pairs and appends it to the PDB file. The atom numbers in these CONECT records are all correct and they work perfectly if the number of atoms is underneath 100000. Specifically, in this case, it is the first line that causes problems. If that one is present, this error is shown:
Bond: maximum coordination exceeded on .R<WAT 30678>.A<H2 3>
-- setting atoms pert=true overrides default limits ATOMS NOT BONDED: .R<WAT 30678>.A<H2 3> .R<CYX 96>.A<SG 8> !FATAL ERROR----------------------------------------
!FATAL: In file [atom.c], line 445
!FATAL: Message: bondAtomProblem found
!
!ABORTING.
followed by a crash to command line. If that one line is omitted, the pdb loads fine.
The SG on CYX96 is atom 1404, which should be bound to the SG on CYX22, which is atom 323. I don't know why it's trying to bond it to H2 on WAT30678, but this is probably due to the faulty pdb. Extracting the pdb with cpptraj (simple trajin restart/trajout pdb) from the final restart file (which contains 100799 atoms, as per header) gives a pdb with 100800 atoms and 30840 residues, whereas the prmtop had 100799 and 30837. This does not occur if the pdb size is underneath 100000. Extracting the pdb with ambpdb gives a pdb with the correct number of atoms and residues, and works fine with those CONECT records.
I can go through the faulty pdb and try to figure out where the extra atom and three extra residues come from, but I was hoping at this point there'd be a solution that doesn't involve me poring over hundreds of thousands of lines of pdb file.
Pietro
________________________________________
From: David A Case [case.biomaps.rutgers.edu]
Sent: 01 December 2014 12:36
To: AMBER Mailing List
Subject: Re: [AMBER] LEaP Error
On Mon, Dec 01, 2014, Aronica, Pietro wrote:
> I'm having problems with LEaP. It has problems loading some perfectly
> fine pdb files, because, I suspect, they have more than 100000 atoms.
> In particular, I think it is misinterpreting the CONECT data of the
> disulphide bridges, trying to link cysteines to water molecules and
> crashing out after having an ATOMS NOT BONDED fatal error.
We need more information than this. There is no particular reason why leap should fail with systems larger than 100000 atoms. Do the CONECT records go beyond that limit? PDB files allocate six digits for the atom number, so generally only have problems with more than a million atoms.
There may well be instances where LEaP messes up parsing of CONECT data.
You will probably have to post a file that illustrates the problem. Be sure that you have converted CYS residues to CYX.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Dec 01 2014 - 14:30:03 PST
