maybe add "entity" (same molecular entity) operator to atomspecs

[Elaine Meng]

A sequence option was recently added to the "select" command and it is very useful. E.g.

select #1 seq /k

...selects all the chains in #1 that are the same biopolymer molecular entity as chain K (which could be in another model or the same model, as I did not specify model after the sequence keyword).

After I tried it a few times I kept thinking how useful it would be as a general atomspec operator, e.g.

color seq /k purple

...color all chains with same sequence as K

[Elaine Meng]

however, in #501 I suggested maybe "entity" would be a better keyword than "sequence"

[Elaine Meng]

related to #1027

[pett]

Is this different in any significant way from #1027? Can we coalesce these two tickets?

[Greg Couch]

Yes, the are essentially the same. Pick your duplicate.

[pett]

This would certainly be nice to have, but it looks like it will be a bear to implement. There is no support for multi-word selectors much less for selectors that take an argument. Modifying the atom specifier syntax itself to support it will be quite the feat, since it isn't a "level" (atom/residue/chain/model) which everything else in the syntax is based around.

I'm open to ideas for any way to get this to happen without a Herculean level of effort...

[pett]

I think I could implement it as a "same_as" or "entity" property of Chain, e.g.:

same_as=/k

Is that too awful?

--Eric

[pett]

Apparently two forward slashes is italics in Wiki formatting. The above was supposed to be:

//same_as=/k

[pett]

Elaine and I like the term "entity" and therefore propose: name this attribute "entity"; add a "byentity" synonym for "bypolymer" to the 'color' command; and remove the "polymer" option from the 'select' command (duplicative with this attribute).

[pett]

I've implemented the mildly hacky entity attribute. It's slightly uglier than my original proposal because you can't used atom-spec special characters directly in an attribute test; you have to quote them. So my original suggestion would instead be:

//entity="/k"

(ignore the leading exclamation point if you're getting this by mail). You can use any spec inside the quotes -- it doesn't have to be a single chain, so it can match (or not match) a set of chains. I'm going to push it out so that people can play around with it and provide feedback.

[Greg Couch]

How does that work when "/k" is composed of multiple (mmCIF) entities? Would you have to do "/k & protein" or "/k & ligand"?

[pett]

Polymeric entities: chains, not ligands.

[Greg Couch]

It would be really great if there were a variation that matched mmCIF entities.

[pett]

I guess I don't understand this comment. This attribute tests whether full sequences match. Don't ligand entities simply have names? Wouldn't you use that?

[Greg Couch]

The point is that mmCIF files have entities, so it would be nice if the concepts matched. Being able to use the mmCIF entity names would be a bonus.

In mmCIF v5, chains could be composed of branched, polymer, and non-polymer entities, all at the same time. It all depends on what the author has picked to use the same chain id.

[pett]

I think you're going down a rabbit hole here. The point of this attribute is to test whether the full primary sequences of chains match. Post-translational modifications like glycosylations, phosphorylations, or covalently bound ligands don't matter for this purpose. Is the problem that the attribute is called "entity"? Would it be better if It were named "sameseq" (or something)?

[Elaine Meng]

I want a word consistent with the coloring by polymer sequence, which is currently "color bypolymer," although I'd originally suggested it should be "color byentity."  Given Greg's aversion to "entity," I vote for either:

(1) polymer (goes with current coloring terminology "color bypolymer")
- or -
(2) sameseq (and we add "color bysameseq" synonym to the existing "color bypolymer").   I also thought of "seqmatch" and "color byseqmatch" but that doesn't seem any better.

Each is yucky/imperfect in its own way, but the ability to specify this concept is useful, and it has to be named something.  Unless anybody else has come up with a brilliant word that solves all our problems.

[pett]

After some mulling things over and discussion with Elaine, we elected to call the attribute "identity". So the new 'color' keyword is "byidentity" (a synonym for "bypolymer") and I have removed the "polymer" keyword from the 'select' command.