Add ncsOperatorList field

[rcsb]

Add ncsOperatorList field

[BobHanson]

Could you elaborate on this? How are these used? Jmol ignores these, I think.

[arose]

Example 1auy:

AU, that is the coordinates explicitly given in the file

trying to construct unitcell without NCS operations

AU expanded by NCS operations

Full biological assembly

Note that in mmCIF files the NCS operations are also in the "XAU" named _pdbx_struct_assembly_gen field.

[arose]

ping @josemduarte

[BobHanson]

Yes, I remember now. This is taken care of in Jmol as

load .... filter "biomolecule XAU"

On Mon, Apr 18, 2016 at 6:11 PM, Alexander Rose [email protected]
wrote:

ping @josemduarte https://github.com/josemduarte

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub
#2 (comment)

Robert M. Hanson
Larson-Anderson Professor of Chemistry
Chair, Department of Chemistry
St. Olaf College
Northfield, MN
http://www.stolaf.edu/people/hansonr

If nature does not answer first what we want,
it is better to take what answer we get.

-- Josiah Willard Gibbs, Lecture XXX, Monday, February 5, 1900

[josemduarte]

An interesting example is 1d2q. It does contain struct_ncs_oper different from an identity matrix, but only of the "given" code, which means the operator can be ignored (see struct_ncs docs in mmcif dictionary):

_struct_ncs_oper.id             1 
_struct_ncs_oper.code           given 
_struct_ncs_oper.details        ? 
_struct_ncs_oper.matrix[1][1]   -1.00000 
_struct_ncs_oper.matrix[1][2]   -0.00164 
_struct_ncs_oper.matrix[1][3]   0.00090 
_struct_ncs_oper.matrix[2][1]   -0.00162 
_struct_ncs_oper.matrix[2][2]   0.99988 
_struct_ncs_oper.matrix[2][3]   0.01547 
_struct_ncs_oper.matrix[3][1]   -0.00092 
_struct_ncs_oper.matrix[3][2]   0.01547 
_struct_ncs_oper.matrix[3][3]   -0.99988 
_struct_ncs_oper.vector[1]      32.84122 
_struct_ncs_oper.vector[2]      18.89500 
_struct_ncs_oper.vector[3]      -32.54992 

That operator seems to be a left over, which wasn't removed after the extra needed chain was added during remediation. Instead of removing it, they decided to assign it a "given" code...

Some more reading about this "interesting" topic:

• a post in pdb-l
• phenix's computational crystallography newsletter from 2014

[BobHanson]

I would like to suggest that this XAU be added to the "bioAssemblyList"
along with the others, and that the transformList include "id" so that "1"
"2" "3" "4" "PAU" and "XAU" are all available. Or whatever is necessary to
reproduce this sort of construction:

loop_
_pdbx_struct_assembly_gen.assembly_id
_pdbx_struct_assembly_gen.oper_expression
_pdbx_struct_assembly_gen.asym_id_list
1 '(1-60)' A,B,C
2 1 A,B,C
3 '(1-5)' A,B,C
4 '(1,2,6,10,23,24)' A,B,C
PAU P A,B,C
XAU '(X0)(1-10,21-25)' A,B,C

Really, it seems to me, X0 itself is not necessary. I do not see the
benefit of adding ncsOperatorList,
as it, like all the other operations (which are also not listed anywhere,
right?), are really just contributions to the biological assembly
transformations. So the more useful thing to do is to create a
bioAssemblyList item that is the set of transformations (X0)(1-10,21-25)
just like all other operations.

Bob Hanson

[BobHanson]

ps -- just noting the statement above "try to depict" the unit cell. One
thing to be clear about is that the depiction given there, also reproduced
below *IS *the crystallized form. That is, the virus capsid protein does
not crystallize in anything like its biological form. So "try" is not
really the right word there. If one is interested in how the structure as
solved in the crystal and how the pieces interact, then that is the right
view to have. Anything else is basically a hypothesis relating to the
actual biological form.

Q: Are there cases where the NCS operator X0 is important?

[image: Inline image 1]

On Mon, Apr 18, 2016 at 7:17 PM, Robert Hanson [email protected] wrote:

I would like to suggest that this XAU be added to the "bioAssemblyList"
along with the others, and that the transformList include "id" so that "1"
"2" "3" "4" "PAU" and "XAU" are all available. Or whatever is necessary to
reproduce this sort of construction:

loop_
_pdbx_struct_assembly_gen.assembly_id
_pdbx_struct_assembly_gen.oper_expression
_pdbx_struct_assembly_gen.asym_id_list
1 '(1-60)' A,B,C
2 1 A,B,C
3 '(1-5)' A,B,C
4 '(1,2,6,10,23,24)' A,B,C
PAU P A,B,C
XAU '(X0)(1-10,21-25)' A,B,C

Really, it seems to me, X0 itself is not necessary. I do not see the
benefit of adding ncsOperatorList,
as it, like all the other operations (which are also not listed anywhere,
right?), are really just contributions to the biological assembly
transformations. So the more useful thing to do is to create a
bioAssemblyList item that is the set of transformations (X0)(1-10,21-25)
just like all other operations.

Bob Hanson

Robert M. Hanson
Larson-Anderson Professor of Chemistry
Chair, Department of Chemistry
St. Olaf College
Northfield, MN
http://www.stolaf.edu/people/hansonr

If nature does not answer first what we want,
it is better to take what answer we get.

-- Josiah Willard Gibbs, Lecture XXX, Monday, February 5, 1900

[arose]

With the try in "try to depict" the unit cell I meant to say that this is not the unitcell because I did not take into account the NCS operations. The image then show a packing that was way to loose to be a crystal. As far as I understand it you construct the unitcell by first applying the NCS operations (when available) to the AU (i.e. the coordinates written in the file) and secondly apply the operations for the given spacegroup to everything constructed in the first step. I maybe wrong.

[arose]

@BobHanson Good point. We currently list all transformations explicitly for each assembly, but it could indeed be useful to do it similar to how it is done in mmcif files. Especially constructs like 1m4x could be represented more concisely.

[josemduarte]

Note that in cases like 1d2q or 1a37 above, there are no XAU bio-assembly given in the files. This shows how the XAU notation was an ad-hoc solution introduced for virus capsid proteins and not used elsewhere.

As such I would vote against recycling the XAU solution for mmtf. We should simply use the extra field ncsOperatorList and represent the data properly: as extra data needed to generate the full crystal. Whilst bioAssembliesList remains a list of one or more interpretations of the possible assemblies present in that crystal (and thus should not include PAU and XAU assemblies, which are not interpretations).

[BobHanson]

Peter, I have a question for you: Is it possible that all the mmCIF files have been normalized so that X0 is always the identity matrix (when it is used in _pdbx_struct_assembly_gen.oper_expression)? I'm not finding any counter-example of that. I'm wondering if that was something that was done as part of the fixing of files some time ago. Jose, I agree. PAU and XAU are specific to viral capsids. They are just subsets of the overall structure. As I understand it, XAU is simply a subset of the symmetries that happens to define a specific pentagonal face in this case. This is marginally useful. I suggest that it is perfect where it is in the mmCIF file. Jmol handles all these the same way, and it works great. There's nothing "special" about XAU other than that it doesn't have a number associated with its ID. Other than that, it is just one more sort of operation. So for 2TBV we have: _pdbx_struct_assembly_gen.assembly_id _pdbx_struct_assembly_gen.oper_expression _pdbx_struct_assembly_gen.asym_id_list 1 '(1-60)' A,B,C,D,E,F,G,H,I 2 1 A,B,C,D,E,F,G,H,I 3 '(1-5)' A,B,C,D,E,F,G,H,I 4 '(1,2,6,10,23,24)' A,B,C,D,E,F,G,H,I PAU P A,B,C,D,E,F,G,H,I XAU '(X0)(1-5)' A,B,C,D,E,F,G,H,I In all these cases, all Jmol needs is the list of 62 non-crystallographic operations -- 1-60, P, and X0. There is no need to duplicate all the matrices for each possible configuration of interest, the way it is done right now. "X0", "1", "2", ... "60" are just character labels pointing to _ pdbx_struct_oper_list.id. It seems to me keeping that simple reference in the mmtf files would be fine. Here "XAU" is valuable because it produces one of the pentagonal faces -- created by operations 1-5 and then oriented by X0 (which is the Identity operation, actually). Similarly "4" is nice because it builds one hexagonal face. This would be a real pain to produce without this guide. Alexander, I realize now that in these cases biomolecular assembly 1 is also the *crystallographic *assembly. I don't know if that is always true, but it is in these cases. To construct the unit cell, one needs to apply the four crystallographic operations, as such: (s1 s2 s3 s4)(1-60) I have never tried that, but I am going to try it now. Bob ​

[arose]

Here is what I get when applying the NCS operations in the 1auy example:

unitcell

supercell (one unitcell in each direction)

[BobHanson]

http://chemapps.stolaf.edu/jmol/zip/jmol-14.5.4_2016.04.19.zip contains Jmol.jar, JmolData.jar, and JSmol JavaScript package for testing.

$ load =1auy.mmtf FILTER "biomolecule 1;.CA,.P"
$ write image clipboard

Inline image 1
FileManager opening url http://mmtf.rcsb.org/full/1JGQ
The Resolver thinks MMTF
MMTF version 0.1
MMTF Producer RCSB-PDB Generator---version: ea7b9e5723271b3f7e46f8a7c5e1cbe18bba3ca2
The Path of Messenger RNA Through the Ribosome. THIS FILE, 1JGQ, CONTAINS THE 30S RIBOSOME SUBUNIT, THREE TRNA, AND MRNA MOLECULES. 50S RIBOSOME SUBUNIT IS IN THE FILE 1GIY
id atoms bonds 1JGQ 8882 5503
1155 ms

$ t =now(); load =1auy.mmtf FILTER "biomolecule 1;.CA,.P";print now(t)
$ write image clipboard

Inline image 2
TURNIP YELLOW MOSAIC VIRUS
701 ms

$ t =now(); load =1auy.mmtf {2 2 1} FILTER "biomolecule 1;bychain";print now(t)
$ write image clipboard
Inline image 3
TURNIP YELLOW MOSAIC VIRUS
595 ms

[BobHanson]

Also, this is up and running in JavaScript at http://chemapps.stolaf.edu/jmol/jsmol/mmtf.htm

Tested only with FireFox/Windows 10. Could be issues because I have switch JSmol to using asynchronous data file loading since these are binary files.

[BobHanson]

This is a test to see if Jmol PNGJ formatted files can be drag-dropped into the Jmol applet or a JSmol JavaScript window on an web page. The image was create using
write PNGJ "virus.png"

[BobHanson]

ps... I have to tell you, I cannot tell at all what I am looking at with
that 1auy supercell image. I'm pretty sure these are *not capsids. *They
are sparse collections of faces. Or some sort of fractional capsids.
Capsids look like the following. (Just showing this as a course-grained
one-particle-per-chain view. Anyway, I think it looks nothing like that.
Did Jmol make that image?

[image: Inline image 2]

On Tue, Apr 19, 2016 at 10:47 AM, Alexander Rose [email protected]
wrote:

Here is what I get when applying the NCS operations in the 1auy example:

unitcell
[image: screenshot 79]
https://cloud.githubusercontent.com/assets/272250/14645504/c45753a2-060a-11e6-8ce9-1ba46ca57949.png

supercell (one unitcell in each direction)
[image: screenshot 78]
https://cloud.githubusercontent.com/assets/272250/14645503/c4575fdc-060a-11e6-88c0-abc9bda24ac7.png

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#2 (comment)

Robert M. Hanson
Larson-Anderson Professor of Chemistry
Chair, Department of Chemistry
St. Olaf College
Northfield, MN
http://www.stolaf.edu/people/hansonr

If nature does not answer first what we want,
it is better to take what answer we get.

-- Josiah Willard Gibbs, Lecture XXX, Monday, February 5, 1900

[arose]

Which image are you refering to? The one below "trying to construct unitcell without NCS operations" or the one below "supercell (one unitcell in each direction)"? The former was wrong because it did not take the NCS operations into account. The images were made with NGL.

[BobHanson]

Sorry - I new the delay would cause confusion. The supercell one. I'm
pretty sure I'm looking at the interior of capsids, right?

On Mon, Apr 25, 2016 at 10:38 AM, Alexander Rose [email protected]
wrote:

Which image are you refering to? The one below "trying to construct
unitcell without NCS operations" or the one below "supercell (one unitcell
in each direction)"? The former was wrong because it did not take the NCS
operations into account. The images were made with NGL
http://arose.github.io/ngl/?load=rcsb://1auy.mmtf.

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#2 (comment)

Robert M. Hanson
Larson-Anderson Professor of Chemistry
Chair, Department of Chemistry
St. Olaf College
Northfield, MN
http://www.stolaf.edu/people/hansonr

If nature does not answer first what we want,
it is better to take what answer we get.

-- Josiah Willard Gibbs, Lecture XXX, Monday, February 5, 1900

[arose]

Yes, you're looking at the interior of capsids. This seems to be how the unitcell is defined (or how I apply the necessary transformations as I don't try to make the complexes as whole as possible).

Here is another perspective, showing that there are full capsid further inside:

[BobHanson]

I must tell you that I can't make much of that. Here is what Jmol produces
for the {1 1 1} cell for 1auy. Now, these are capsids. They are created
by the command:

load =1auy.mmtf {1 1 1} filter "biomolecule 1;*.ca"

Which applies the full 60 non-crystallographic symmetries to get the
capsid, then, after that, applies the space group operations of group 181
to place and orient those capsids in the unit cell.

[image: Inline image 1]

On Mon, Apr 25, 2016 at 1:28 PM, Alexander Rose [email protected]
wrote:

Yes, you're looking at the interior of capsids. This seems to be how the
unitcell is defined (or how I apply the necessary transformations as I
don't try to make the complexes as whole as possible).

Here is another perspective, showing that there are full capsid further
inside:

[image: screenshot 96]
https://cloud.githubusercontent.com/assets/272250/14794419/d1c376ce-0ad8-11e6-9044-8d0a4825d325.png

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#2 (comment)

Robert M. Hanson
Larson-Anderson Professor of Chemistry
Chair, Department of Chemistry
St. Olaf College
Northfield, MN
http://www.stolaf.edu/people/hansonr

If nature does not answer first what we want,
it is better to take what answer we get.

-- Josiah Willard Gibbs, Lecture XXX, Monday, February 5, 1900

[arose]

What do you mean by "60 non-crystallographic symmetries"? I mean the data that is in the struct_ncs_oper mmcif field. Note that this field is currently not included in the mmtf format, so I am loading cif in the examples with NGL. I apply struct_ncs_oper operations and then those from group P 64 2 2 (which I guess is 181). After that I shift the unitcell in each direction to create the supercell.

[BobHanson]

I agree that we need the struct_ncs_oper block and that is one way to do this. Please do add those. I'm probably getting away with this because in these viral cases the biological unit is also in the crystal. (That's not always the case, right?) My picture is "prettier" because it organizes the visualization along the lines of the biologically relevant unit. So I suspect both of our representations are correct. What's happening here is that there are three viral particles per unit cell. Each unit cell has 12 copies of the basic unit due to crystallographic operations. So that means that each operation contributes four units to each viral particle. Thus, in order to get 60 units per viral particle you need to add an additional 15 operations. These are your added NCS operations. Do you know of some specific cases where the biological assembly is not actually present in the crystal structure? It is annoying that the PDB file lists the crystallographic operations and the biomolecule operations (mixing crystallographic and noncrystallographic); the mmCIF file lists the noncrystallographic operations, but not the crystallographic ones; and the mmtf (as of now) lists the biomolecule operations only. None of them have the information in a standard CIF file that allows full construction of the actual crystal: a simple list of operators! But I think the PDB file comes closest, in that it does list the Jones-Faithful (x-1/2, y, z) notation for the space group, at least -- acknowledging the fact that one cannot just read a space group name and expect to get the operators right.

[BobHanson]

here's an image from Jmol in the orientation of the first "trying to
construct unitcell without NCS operations" image in this thread. From it
you can see (sort of) how those sparse units fill in to give full capsids.
It's not a great perspective.

[image: Inline image 1]
Here's one that shows how four basic crystallographic units fit in with the
units on a single capsid:
[image: Inline image 2]

So these capsids must be hiding somewhere in your image there.

​

[arose]

I agree that we need the struct_ncs_oper block and that is one way to do this. Please do add those.

We will!

My picture is "prettier" because it organizes the visualization along the lines of the biologically relevant unit.

Yeah, it is. I should do that to.

Do you know of some specific cases where the biological assembly is not actually present in the crystal structure?

@josemduarte you said no, right? At least it shouldn't.

@BobHanson the images you send to the github issue thread don't show up.

[BobHanson]

OK, I see. I have to come here and enter my message, not do that from a gmail replay.

For Jmol I have added the NCS operations for when there is no request for a biomolecule but there is a request for a unit cell:

load =1auy.cif {1 1 1}

but not

load =1auy.cif {1 1 1} filter "biomolecule 1;*.CA"

Result is a bit ragged.

Thanks very much for clearing up for me the use of NCS operations.

[josemduarte]

To be clear: in the current definition the biological assembly must be a part of the crystal. The bio assembly annotation is simply an interpretation of what is a plausible assembly in the crystal. Usually that interpretation should come from independent experimental data confirming the oligomeric state in solution.

[BobHanson]

Good to hear. Thank you.

[BobHanson]

Is that enhancement pushed out to http://mmtf.rcsb.org ?

[arose]

No, that will take some time, while we are gathering feedback. We will let you know!

[josemduarte]

This has been done in 0.2 already, hasn't it? Shall we close?

[arose]

documented in spec https://github.com/rcsb/mmtf/blob/master/spec.md#ncsoperatorlist