Amber mask syntax for chainid selections?

[jchodera]

Is there some special syntax for Amber mask selection of specific chain IDs?

[jchodera]

It looks like UCSF Chimera does support chain selections in the syntax:
https://www.cgl.ucsf.edu/chimera/docs/UsersGuide/midas/atom_spec.html

I guess this comes down to the question of whether the ParmEd "very similar to" syntax includes chain support.

[jchodera]

From looking at mask.py, looks like maybe not yet, but perhaps it could be extended to include chain selections?

[swails]

Amber never traditionally stored chain information, so it was never supported in the mask syntax. Have a look at the selection/slicing syntax supported by Structure: http://parmed.github.io/ParmEd/html/structure.html#structure-manipulation-slicing-combining-replicating-and-splitting

Selecting a particular chain is pretty easy. Let's say you want chain A:

import parmed as pmd
struct = pmd.load_file('something.pdb')
chainA = struct['A',:,:] # Select all residues in chain A and all atoms in that residue
chainAB = struct[['A', 'B'],:,:]

Each will generate a new instance of type(struct) that contains each of the atoms and valence terms/parameters present in the selected region (from which you can simply access the topology attribute for an OpenMM Topology.

[swails]

You can also generate a pandas DataFrame from the system and use very simple expressions to generate a mask array to feed to the Structure selector, if you're more comfortable with pandas...

[hainm]

@swails is there intention to just chainA = struct['A']? or may be chainA = struct['chain A']?

may be add a method named select to make the selection more explicit?

[swails]

@swails is there intention to just chainA = struct['A']?

No. struct['A'] interprets 'A' as an Amber mask. In order to have it recognized as a chain, you need to do struct[[<chains>:[<residues>:]]<atoms>] (where <chains> and <residues> are optional).

or may be chainA = struct['chain A']?

No. If I introduced a syntax like that, I'd either do what MDTraj does (use pyparsing to reduce selection strings to Python expressions) or implement it in a select method that was separate from the [] syntax.

may be add a method named select to make the selection more explicit?

This is a possible extension, but not one that is important to me.

[drroe]

@swails I've been thinking of expanding the Amber mask parser to be able to
do selections like this (by chain ID, molecule #, etc), but we should
probably talk about what operator characters to use first.

On Mon, Jun 22, 2015 at 10:58 AM, Jason Swails notifications@github.com
wrote:

@swails https://github.com/swails is there intention to just chainA =
struct['A']?

No. struct['A'] interprets 'A' as an Amber mask. In order to have it
recognized as a chain, you need to do
struct[[:[:]]](where and
are optional).

or may be chainA = struct['chain A']?

No. If I introduced a syntax like that, I'd either do what MDTraj does
(use pyparsing to reduce selection strings to Python expressions) or
implement it in a select method that was separate from the [] syntax.

may be add a method named select to make the selection more explicit?

This is a possible extension, but not one that is important to me.

—
Reply to this email directly or view it on GitHub
#209 (comment).

---

Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 307
Salt Lake City, UT 84112-5820
http://home.chpc.utah.edu/~cheatham/
(801) 587-9652
(801) 585-6208 (Fax)

[hainm]

+1 for this. I've been waiting.

[swails]

A couple ideas:

1. Do what Chimera does (see https://www.cgl.ucsf.edu/chimera/docs/UsersGuide/midas/frameatom_spec.html)
2. Use :: for chain IDs (in the same way that @% indicates atom type names)

An example of 1:

# Select residues 1-10 in chain A
:1-10.A
# Select residues 1-10 in chains A, B, and C
:1-10.A-C # I *think*; not sure if you can use - for string range, maybe .A,B,C
# etc...

An example of 2:

# Select chain A
::A
# Select residues 1-10 of chains A, B, and C
::A,B,C:1-10
# Select protein backbone atoms of residues 2-20 in chains A and B
::A,B:2-20@CA,C,O,N

Another idea for flexibility is to say that chains are strings, and use the same "chain" syntax to denote molecules when the chain is given a numerical value.

[jchodera]

I'm not sure we even finished this discussion! I think either format works for us!

[swails]

This isn't the right place to talk about expanding the Amber mask syntax. There are (at least) 3 mask parsers that will need to be updated for this -- the one in cpptraj, the one in sander/pmemd, and the one in ParmEd. The one in sander/pmemd is probably less important, unless we want to start having sander/pmemd understand the optional CHAINID flag that ParmEd's addPDB flag can add. But that is almost certainly not going to happen (lots of work for minimal gain).

But this will require a discussion/decision from the Amber dev team. I'll bring it up at the developer's meeting. But I'm closing this here.