Change ChEBI xref used for Fru(6)P (R-ALL-29512.5) in Glycolysis/Gluconeogeneis/Pentose Phosphate Pathway

[sjm41]

In Glycolysis (R-HSA-70171) and Gluconeogeneis (R-HSA-70263):

• Change the xref on Fru(6)P (R-ALL-29512.5) from D-fructofuranose 6-phosphate(2−) CHEBI:61527 to β-D-fructofuranose 6-phosphate(2−) CHEBI:57634

Same change In the Pentose Phosphate Pathway (R-HSA-71336):

• change xref on Fru(6)P (R-ALL-29512.5) from D-fructofuranose 6-phosphate(2−) CHEBI:61527 to β-D-fructofuranose 6-phosphate(2−) CHEBI:57634

[deustp01]

I have a problem. Fru(6)P (R-ALL-29512.5) (D-fructofuranose 6-phosphate(2−) CHEBI:61527 is used in a really large and varied set of reactions. Is there any way of knowing whether the suggested replacement, β-D-fructofuranose 6-phosphate(2−) CHEBI:57634, is correct for all of them?

It may well be that the beta form is universally correct for mammals - this is a question out of ignorance. Any clues as to the answer?

If in fact the proposed replacement - use CHEBI:57634 instead of CHEBI:61527 - is valid everywhere (or even plausible in the absence of evidence to the contrary) then it is really easy to do the swap and in addition all connectivity between reactions involving fructose-6-phosphate is preserved. Easy! On the other hand, if different reactions are likely to use different stereoisomers / anomers / whatever then the cleanup will require hard detailed expert manual curation and connectivity among reactions that are now linked because one generates fructose-6-phosphate and another consumes it or is regulated by it will likewise need to be manually reworked.

Help!

[sjm41]

Just to clarify: is it not possible just to change the instance of Fru(6)P in Glycolysis (R-HSA-70171) and Gluconeogeneis (R-HSA-70263) in Reactome? That is, if you change it there, you have to change it everywhere?

In terms of your question (is the beta form is universally correct for mammals) - can you provide the RHEA IDs associated with each of the Reactome reactions involving Fru(6)P?
We could then see if RHEA, at least, says if all of these reactions use the beta form.

Maybe @ukemi or @rozaru also have insights?

[ukemi]

I don't know the biochemistry, but I think if Reactome were to just make the chemical more specific for an individual reaction, then Reactome would not consider it the same chemical used elsewhere. This is a bit unsatisfying. Does it hurt to universally use a more generic parent here? I suppose it would make the models be different levels of granularity if curators cloned and changed. How would this effect things? It depends on how we compute on the models at the end of the day.

[sjm41]

I think if Reactome were to just make the chemical more specific for an individual reaction, then Reactome would not consider it the same chemical used elsewhere. This is a bit unsatisfying.
Ah, I see.

Does it hurt to universally use a more generic parent here?
No, it's fine. I was just pointing out the slight inconsistency with MetaCyc/RHEA for this particular reaction in case it could be made more specific - it's good to be consistent where possible!

If all the RHEAs associated/xreffed to the Fru(6)P reactions in Reactome are specific the beta form, then I'd say there's a reasonable argument for making the wholesale change in Reactome. If some RHEAs specify the alpha/agnostic form, then Reactome should definitely stick with the more generic parent. But that's Peter's call in the end!

[deustp01]

There are issues of consistency and granularity for annotation, that feed into issues of data consistency for data mining, that we need to sort out here. Thinking at the level of a cartoon, I think I want to represent the fact that an output of reaction A is an input of reaction B, and by that I mean that the identical physical entity in all its ionized and stereochemical detail is the output / input, not just a member of a small family of closely related entities. Reactome process description diagrams require this and, if I understand our previous discussions, so does "directly provides input for" reasoning in GO-CAMs.

On the other hand, requiring this chemical identity may go beyond the actual data, or may ignore additional reaction steps - maybe an output must spontaneously isomerize or ionize to become a workable substrate for the following reaction, and this spontaneous event, even if thermodynamically unfavorable, happens to a sufficient step to allow the downstream reaction to pull the unfavorable spontaneous one in a forward direction.

There are probably use cases enough for a long weeds discussion in reactions involving "fructose", at least as complicated and painful as my attempt to find an exact stereochemical pathway through the early part of glycolysis several years ago.

RHEA_glycodisconnects.docx

This certainly doesn't scale and is often probably not really justified by available evidence. But then we need some agreement on what level of granularity to use (as in the alternatives suggested in this ticket) among ourselves and RHEA and GO. (Mostly, I expect, with RHEA, figuring that GO will follow RHEA).

Final metaphysical thought - this looks like a discussion of how to apply the notion of broad / narrow / exact synonyms to chemical entities and families of them consistently across GO and RHEA and Reactome and ChEBI.

Final practical thought - despite all this argument, the only urgent need I see is to keep directly-provides-input-for causal connections between successive Reactome reactions / derived GO-CAM activity units, in a way that minimized legacy cleanup and supports reliable annotation and maintenance by future curators. Getting all the anomeric carbons right is a chore I would happily skip for myself and not inflict on anyone else, if we can figure out how to do it consistently and without loss of important information.

[sjm41]

I think it's absolutely fine for Reactome to continue to use the anomer-agnostic "D-fructofuranose 6-phosphate(2−) CHEBI:61527" for Fru(6)P.
In fact, now I understand that the Fru(6)P identity has to fit ALL the associated reactions in Reactome, and that there are many such instances, this seems the only feasible approach!

However, for GO-CAMs, it's possible (and it seems appropriate) to use an anomer-specific form for a specific reaction if there is evidence to do so, and/or if the RHEA associated with the given GO term uses a specific anomer.
That appears to be the case for the reactions involving Fru(6)P in Glycolysis, Gluconeogeneis and Pentose Phosphate Pathway. That is, RHEA specifies the beta anomer for those particular reactions, so I think the GO-CAMs should continue to use the beta anomer. (This is all assuming RHEA has got things right, and that the GO term has the correct RHEA xref....)

Of course, this does cause a small issue for the Reactome -> GO-CAM imports, in that they will get the generic CHEBI:61527 for Fru(6)P in the 3 pathways mentioned above, whereas participants derived from the associated GO terms will get the specific beta anomer CHEBI:57634.

P.S. I think the disconnects you previously identified with glycolysis are now mostly OK, in that a legal GO-CAM can be built with contiguous input/output chemicals (e.g. see Rossana's 65a1f4f800000617 model) that is the same as the Reactome model, except for the Fru(6)P discrepancy mentioned above.
See my analysis here: https://docs.google.com/spreadsheets/d/1ApvzyHRO9xwX-yOc0nmPest29tm6DqvYf4dwhv5BqcI/edit?gid=0#gid=0