definition vs intended use of GO:0030170

[gocentral]

I'm updating P00508, a protein with covalently-bound pyridoxal phosphate as cofactor (see PMID:1593633)

How to convey that this enzyme has a covalently bound pyridoxal phosphate as cofactor? According to the definition (http://www.ebi.ac.uk/QuickGO/GTerm?id=GO:0030170) this term should not be used when a protein interacts covalently with pyridoxal phosphate.

Reported by: uhinz

Original Ticket: geneontology/annotation-issues/735

[gocentral]

Hello,

Your interpretation of the definition and intended usage of GO:0030170 is correct. Because GO is meant to capture activities and roles of gene products, it doesn't really support annotating physical features of a protein such as sequence, folding, or covalently bound cofactors.

-Midori

Original comment by: mah11

[gocentral]

My confusion stems from the fact that this GO term is widely applied for proteins with covalently bound pyridoxal phosphate (IEA, based on InterPro or HAMAP) (see http://www.ebi.ac.uk/QuickGO/GTerm?id=GO:0030170)

Could something be done to clean up the electronic annotation?

Original comment by: uhinz

[gocentral]

assigning to one of the GOA annotators to see if the problem can be addressed

Original comment by: mah11

[gocentral]

• assigned_to: nobody --> edimmer

Original comment by: mah11

[gocentral]

But pridoxal phosphate is often covalently bound. Would it break anything else in the function ontology simply to delete the "and covalently" phrase from the definition? Indeed, to propagate that deletion all the way up to the ancestral term GO:0019842 vitamin binding, so the definition would read "Interacting selectively with a vitamin, one of a number of unrelated organic substances that occur in many foods in small amounts and that are necessary in trace amounts for the normal metabolic functioning of the body." The formation of a stable, specific association between vitamin cofactor and apoprotein is functionally critical; the exact chemical forces (covalent, noncovalent) mediating the stability can vary from case to case as long as they exist.

Original comment by: deustp01

[gocentral]

I completely agree with Peter. Cofactor binding was decided by the GOC to be a valid use for such terms. It would be a loss to only be able to annotate a subset of the proteins that bound cofactor pridoxal phosphate.

Original comment by: edimmer

[gocentral]

Personally, I'm neutral, but when the issue arose earlier, some curators argued strongly for the exclusion of all covalent bonding from the 'binding' definition. These are the relevant threads on the annotation mailing list:

http://fafner.stanford.edu/pipermail/annotation/2008-January/001356.html http://fafner.stanford.edu/pipermail/annotation/2008-January/001363.html http://fafner.stanford.edu/pipermail/annotation/2008-January/001367.html

I think Emily plans to get in touch with Harold to raise the question again in light of this item.

m

Original comment by: mah11

[gocentral]

The thread Midori cited is clear: in GO, binding must be noncovalent. A problem is that our biochemist users cannot always distinguish noncovalent binding from covalent potrzebing (where potrzebi is a placeholder for the word to be invented to describe the covalent association of a cofactor like pyridoxal phosphate with an enzyme).

Here's the Devlin textbook (5th ed, 2002, Wiley - page 783) on the subject: "The active site of the 'resting' aminotransferase contains pyridoxal phosphate covalently attached to an epsilon-amino group of a lysine residue that forms part of the primary structure of the transferase ... . When a substrate amino acid ... approaches the active site, its amino group displaces the lysine epsilon-amino group ... . At this point the pyridoxal phosphate-derived molecule is no longer covalently attached to the enzyme ..." The passage goes on to describe additional re-bonding gymnastics whereby the amino acid becomes a ketoacid, a ketoacid becomes an amino acid, and the pyridoxal phosphate (always bound, in the GO sense) is finally restored to its "resting" state covalently potrzebied to the enzyme.

The pyridoxal cofactor functions by switching between states of covalent and noncovalent attachment to the aminotransferase enzyme and to use the current definition of "binding" in GO correctly, we can't make any legal molecular function annotation concerning the relationship between an aminotransferase and its cofactor except in the special case of a reaction mechanism study that has precisely specified a step in the multistep catalyzed reaction and demonstrated the exact bonding structure of the cofactor, substrate and enzyme protein at that step.

Peter

Original comment by: deustp01

[gocentral]

I feel strongly that the term binding remain a non-covalent interaction. Binding infers Ka and Kd (reversibility). You will find this in many chemistry and biochemistry textbooks; Given the gymnastics required to get it back on, I would not classify this as "binding".

Original comment by: hdrabkin

[gocentral]

Where we come out, perhaps, is that an additional term is needed. Perhaps "association" could be a parent that has "non-covalent binding" (exactly equal to GO "binding" in its current correct usage) and "covalent binding" (a new term) as children. That also allows use of the parent for cases like pyridoxal phosphate where the interaction is too complex to be captured accurately by either term. (Perhaps someone can come up with a better parent and pair of child terms.)

There's also a granularity-of-annotation problem looming here. If we are to be compelled to distinguish covalent from non-covalent interactions then correct annotation of something like that enzyme - pyridoxal phosphate cofactor interaction can no longer be captured in a single term but instead requires at least two and in many cases cannot be captured at all because the data in the paper being annotated only tell us that the cofactor must be associated with the enzyme in order for the enzyme to function - the experimenter didn't do the additional hard work to determine the atomic details of the interaction.

As a minor point, I'm hesitant to use Km / Kd / reversibility criteria as a defining feature to distinguish covalent and non-covalent interactions. As those same textbooks point out, although one hydrogen bond (a non-covalent interaction) is weak, the cumulative effect of many of them can often mediate an interaction that is effectively irreversible under physiological conditions.

Original comment by: deustp01

[gocentral]

What I am afraid of is that then "methyl binding" will be applied to a protein getting methylated, etc. I think that if two things covalently bind each other it means a covalent bond is formed between them. In cases where PP is covalently attached to a protein to act as it's cofactor, is the attachment mediated by an additional enzyme or does it occur by virtue of a property of the protein that it is attached to itself?

Original comment by: hdrabkin

[gocentral]

Agreed - cofactor association of whatever sort needs to be distinguished from covalent posttranslational modification. Could this be handled with a comment? I think that we could all look at any given protein - small molecule association and classify it as cofactor association versus covalent posttranslational modification (though I'm at a loss for the words to turn that hunch into a proper definition and test).

Original comment by: deustp01

[gocentral]

• status: open --> closed

Original comment by: edimmer