Atoms lost reading pdbe_bio file, atoms not grouped by model in mmCIF file

[pett]

open pdbe_bio:2MIV maxAssemblies 1 autoStyle false

loses the nucleic acid part of the structure (which is clearly in the file)

[Tom Goddard]

I've attached the mmcif file. It contains 5 models (from NMR). The atoms for each model are not grouped together in the atom_site table. The atom order in the table is 1A (model 1 chain A), 2A,3A,4A,5A, 1B,2B,3B,4B,5B, 1C,2C,3C,4C,5C. So all the chain A atoms for the 5 models come before the chain B atoms. The only atoms that get opened by ChimeraX are in chain C.

The mmCIF reader should warn when the atoms are in an illegal order, if in fact this order is illegal.

[Tom Goddard]

2miv assembly 1 from PDBe

[Tom Goddard]

I've asked the PDB to clarify if the atom records in the mmCIF atom_site table have to be grouped together for each model.

From: Tom Goddard 
Subject: mmCIF atom_site table line order
Date: December 8, 2022 at 4:16:13 PM PST
To: pdbx-mmcif-help@mail.wwpdb.org

Is the ordering of atoms in the mmCIF atom_site table constrained?  For instance do all atoms of the same model have to be contiguous in the atom_site table?  Or is any order allowed for the atoms in the table?  I ask because the PDBe biological assembly file 2miv which contains 5 models has atoms from the different models interleaved.  That causes the ChimeraX mmCIF reader to fail because it assumes that atoms with the same pdbx_PDB_model_num in the atom_site table will all be grouped together.

I've attached the PDBe file if you are interested to see this case with atoms from different models not placed contiguously in the atom_site table.

I was not able to find any statement about the order of atoms in the atom_site table in the online documentation of mmCIF.

Thanks!

	Tom Goddard
	ChimeraX developer

[Tom Goddard]

Reply from PDB mmCIF support -- they are looking into it.

From: Ezra Peisach 
Subject: Re: mmCIF atom_site table line order
Date: December 8, 2022 at 6:36:02 PM PST
To: Tom Goddard <goddard@sonic.net>

I will need to discuss internally.

A year or so ago, the wwPDB came up with a standard assembly cif file - but it appears the PDBe website is not offering up this file.

Intuitively, mmCIF follows the same type of layout as PDB - one would not expect model numbers to jump around.

It also makes no sense here - as what is an "assembly file" for a multi-model file. You take your reference model (usually model 1).


Ezra

[Greg Couch]

"fixed" this by extending previous found PDB (NMR) model. You will now get a warning "Previously completed PDB model 1 found. Trying to extend." It should always succeed at extending it, but I don't want to make any promises. You'll also see a warning about a missing entity_poly_seq table. It is a bad mmCIF file in many ways.