trr: propose to change parent for 'chromatin binding'

[gocentral]

I propose to change the the parent of GO:0003682 : chromatin binding from GO:0003677 : DNA binding to GO:0005488 : binding

as chromatin binding is not a type of DNA binding. In fact, it is the other way round, DNA binding IS A type of chromatin binding. I.e.
GO:0003677 : DNA binding could (also) be made child of
GO:0003682 : chromatin binding in addition to being child of GO:0003676 : nucleic acid binding.

I have no reference for this, as it seems rather obvious!

RAa

Reported by: rein_aa

Original Ticket: "geneontology/ontology-requests/2574":https://sourceforge.net/p/geneontology/ontology-requests/2574

[gocentral]

Logged In: YES user_id=436423

I agree that chromatin binding should not be is_a DNA binding because chromatin contains protein as well as DNA. DNA binding cannot be made a child of chromatin binding, however, because it refers to physical interaction with any DNA, including forms that are not assembled into chromatin.

We could therefore move GO:0003682 up to be a direct child of GO:005488, but leave the parentage of GO:003677 unchanged.

Midori

Original comment by: mah11

[gocentral]

Logged In: YES user_id=1294488

I see your point, Midori, and it would certainly help if GO:0003682 got mooved up.

BUT, I thought that 'GO:0003677 : DNA binding' could have TWO parents ('nucleic acid binding' (GO:0003676) AND 'chromatin binding'), which would reflect "two meanings" of DNA binding (given it's parent), one for binding to any DNA, the other for DNA binding in the context of chromatin. As I understand the idea of multiple parenthood, I thought this was exactly the kind of case where this would be corrrect. Compare e.g. the case for 'receptor binding' (GO:0005102) which is child of both 'protein binding' (GO:0005515) and 'signal transducer activity' (GO:0004871).

The alternative could be to make a new entry for DNA binding sensu chromatin (and/or nucleosome).

BUT, I realise that I may not have fully understood the ideas behind multiple parenthood.

Rein

Original comment by: rein_aa

[gocentral]

Logged In: YES user_id=436423

Multiple parentage implies that both or all parents are always true, not that one or another may be true. So if DNA binding were made a child of chromatin binding, that would imply that DNA binding is always chromatin binding.

We also call this the 'true path rule', as in the documentation on the GO web site:

http://www.geneontology.org/GO.usage.shtml?all\#truePathRule

One of the main reasons for having this rule is for gene product annotations: a gene product annotated to any term is implicitly annotated to every parent along every path back to the root of the ontology.

We can add a new term for 'chromatin DNA binding', which would be a child of both chromatin binding and DNA binding. I'll send a message to the GO mailing list first.

Midori

Original comment by: mah11

[gocentral]

Logged In: YES user_id=546388

Most definitely, chromatin binding cannot be a direct child of DNA binding, since anything binding chromatin could be doing so by binding to protein, DNA, or both. Direct parentage to "binding' is warrented.

Original comment by: hdrabkin

[gocentral]

Logged In: YES user_id=1294488

Thanks for educating me on the true-path-rule! I should have known!

'Chromatin DNA binding' as child of 'chromatin binding' and 'DNA binding' would be good.
Similarly, if my proposed term: nucleosome binding is accepted, it could get as child 'nucleosomal DNA binding'.

cheers

Rein

Original comment by: rein_aa

[gocentral]

Logged In: YES user_id=436423

I've just sent this suggestion to the GO mailing list (where there has been a bit of discussion):

Here's my latest proposal:

binding ; GO:0005488 --[i] DNA binding ; GO:0003677 ----[i] chromatin DNA binding ; GO:new1 --[i] chromatin binding ; GO:0003682 ----[i] nucleosome binding ; GO:new2 ------[i] chromatin DNA binding ; GO:new1 ------[i] chromatin histone binding ; GO:new3 --[i] protein binding ; GO:0005515 ----[i] histone binding ; GO:0042393 ------[i] chromatin histone binding ; GO:new3

Any comments?

Is there a need for a term for 'free histone binding', meaning binding histones that are not assembled into chromatin? Or will annotating to GO:0042393 suffice? (I will only add this term if it is explicitly requested).

m

Original comment by: mah11

[gocentral]

Logged In: YES user_id=1294488

'Histone binding' should be fine for binding to free histones, and many of the proteins that do this (e.g. histone chaperones and nucleosome assembly proteins) are already annotated with this term.

A chromatin afficionado might, however, not like the proposed chromatin binding subtree. As it is, it says thatall chromatin DNA binding is also nucleosome binding. That is not the case as LINKER DNA between the nucleosomes proper is stricktly speaking not nucleosomal DNA. The same is true for nucleosome-free regions in chromatin. So the thrue path rule is violated as there is chromatin DNA that is not nucleosomal.

What about this:

binding ; GO:0005488 --[i] DNA binding ; GO:0003677 ----[i] chromatin DNA binding ; GO:new1 --[i] chromatin binding ; GO:0003682 ----[i] nucleosome binding ; GO:new2 ------[i] nucleosomal DNA binding ; GO:new1 ------[i] nucleosomal histone binding ; GO:new3 ----[i] chromatin DNA binding ; GO:0003682 --[i] protein binding ; GO:0005515 ----[i] histone binding ; GO:0042393 ------[i] nucleosomal histone binding ; GO:new3

If I managed to keep the head clear now!

Rein

Original comment by: rein_aa

[gocentral]

Logged In: YES user_id=1294488

... sorry .... got one GO id wrong there: chromatin DNA binding ; GO:0003682 should be chromatin DNA binding ; new1

Rein

Original comment by: rein_aa

[gocentral]

Logged In: YES user_id=436423

I'm embarrassed to admit it, but I'd forgotten about linker DNA when I drew up the proposal! Someone on the mailing list noticed as well. Your revised structure will do (assuming corrections to new GO IDs), and can include this relationship as well:

----[i] chromatin DNA binding ; GO:new2 ------[i] nucleosomal DNA binding ; GO:new3 (or 4 or ...)

Original comment by: mah11

[gocentral]

Logged In: YES user_id=1294488

Now it is looking pretty good! I think we have done a good job here :-)

Rein

binding ; GO:0005488 --[i] DNA binding ; GO:0003677 ----[i] chromatin DNA binding ; GO:new1 ------[i] nucleosomal DNA binding ; GO:new2 --[i] chromatin binding ; GO:0003682 ----[i] nucleosome binding ; GO:new2 ------[i] nucleosomal DNA binding ; GO:new2 ------[i] nucleosomal histone binding ; GO:new3 ----[i] chromatin DNA binding ; GO:new1 --[i] protein binding ; GO:0005515 ----[i] histone binding ; GO:0042393 ------[i] nucleosomal histone binding ; GO:new3

Original comment by: rein_aa

[gocentral]

Logged In: YES user_id=436423

OK, I've implemented the latest structure; IDs for new terms are:

chromatin DNA binding GO:0031490 nucleosome binding GO:0031491 nucleosomal DNA binding GO:0031492 nucleosomal histone binding GO:0031493

Original comment by: mah11

[gocentral]

• assigned_to: nobody --> gomidori
• status: open --> closed-accepted

Original comment by: mah11