Pathwaybot

cites work (P2860)

In connection with a pull request to Scholia, I come to wonder why biological pathways use cites work (P2860) instead of described by source (P1343)? — Finn Årup Nielsen (fnielsen) (talk) 16:15, 1 August 2017 (UTC)

This is in relation to https://github.com/fnielsen/scholia/pull/174 — Finn Årup Nielsen (fnielsen) (talk) 16:18, 1 August 2017 (UTC)

@Fnielsen: Both the the Wikipathways and Reactome records, which are being sourced here are pathway descriptions and not so much pathways themselves. So the related Wikidata entries a replication of those Wikipathways and Reactome descriptions, and it is the description that cites a specific citation. Also, a citation does not necessarily describe the pathway under scrutiny, but only a very specific part of that pathway. --Andrawaag (talk) 18:26, 1 August 2017 (UTC)

@Andrawaag: But if the items are "pathway descriptions" descriptions and not "pathways", then they should be instance of "biological pathway description" and not "biological pathway". Also the label and description should reflect that I think? — Finn Årup Nielsen (fnielsen) (talk) 19:21, 1 August 2017 (UTC)

@fnielsen: That is a very good point. I might agree with you on adding that as a description, with respect to adding that as an instance of a pathway description, that is a bit more complicated, since those P31 values in the near future point to items sources from ontologies. On second thoughts, biological pathways are abstractions of biological processes, so we could also argue that pathways them selves are descriptions of that process. Going back to the question if described by source, would be more appropriate, I still don't think so. Described by source indicates some level of completeness, which often isn't the case. A cited source, often only describes a very specific part of a pathway. --Andrawaag (talk) 19:32, 1 August 2017 (UTC)

(edit conflict)

I see that tricarboxylic acid cycle (Q133895) (a "real" pathway) has described by source (P1343) to the pathway descriptions, both one for Reactome and one for Wikipathway. — Finn Årup Nielsen (fnielsen) (talk) 19:34, 1 August 2017 (UTC)

Indeed, but that is the process (with its proper Wikipedia entry), which has descriptions in WIkipathways and Reactome. So here I would argue that the use of described by source (P1343) is appropriate. --Andrawaag (talk) 19:37, 1 August 2017 (UTC)

But then you have, e.g., Q28031601 which has part(s) (P527) phosphate(3−) ion (Q177811). How can a description consist of phosphate? I does not make sense to me. — Finn Årup Nielsen (fnielsen) (talk) 19:40, 1 August 2017 (UTC)

@fnielsen: that record was a first modelling effort of Reactome pathways and indeed no longer makes sense. I have just requested its deletion. --Andrawaag (talk) 19:56, 1 August 2017 (UTC)

A thought I had about the use of 'cites'. It is the generally accepted way of referring to a publication. The list of references in a paper does not try to categorise the entries, even though they could probably be qualified with 'papers I just mention'; 'paper I used data/ideas from'; etc ... I can imagine a wikidata query that looked for any items that cite a particular paper; finding pathways that were derived/deduced from that paper would be valuable information. If we did not use 'cites' we put the onus on the user to work out what property we did use to connect a pathway to a publication. SarahKeating

Diseases

Hello,

how could we link Disease (Q45314346) and disease (Q12136) ? (and child items like Diseases of metabolism (Q45316399) and metabolic disease (Q2351083) )

maybe using has effect (P1542) ?

Disease (Q45314346)

has effect (P1542) disease (Q12136)

this work well for leaf nodes

Defective SLC26A2 causes chondrodysplasias (Q45314499)

has effect (P1542) chondrodysplasia (Q26271752)

genetic association (P2293) SLC26A2 (Q15311810)

it's even debatable whether Defective SLC26A2 causes chondrodysplasias (Q45314499) could be merged in some way into chondrodysplasia (Q26271752)

Ske (talk) 10:09, 8 May 2018 (UTC)

Q143623

This [1] edit was not correct in any way. Wikidata items are not property of Reactome to be reduced to such data, maybe something what you call Defined set from Reactome should be different items from items describing general concepts in chemistry. Description change was wrong, deleting existing P31 was wrong, using has part(s) (P527) for members of compound class was wrong. Please be more cautious. Wostr (talk) 17:41, 2 October 2018 (UTC)

In fact these two items were separated in the past, so probably you should undo the merge? Wostr (talk) 17:46, 2 October 2018 (UTC)

taxon of Reactome import

Unfortunately you cannot assume that all entities from Reactome are from human organism. See User:SCIdude/Protein_bugs#Reactome_R-MTU_imports_marked_as_human. BUt these are only MTU, they also have Listeria, HIV, Influenza entities. Please fix. --SCIdude (talk) 08:23, 1 October 2019 (UTC)

Incorrect modeling

@AlexanderPico, Andrawaag, Egon Willighagen, Gstupp: The pathway itself is not the same as articles describing them (including entries in WikiPathways). A pathway does not have an author. Nor it can cite any article. You should create separate items for articles describing a pathway and link them using main subject (P921) and described by source (P1343).--GZWDer (talk) 03:48, 13 February 2020 (UTC)

If a pathway is a description of a process, I don't think has part(s) (P527) (about gene) is appropriate. Why a gene is a part of a work?--GZWDer (talk) 04:10, 13 February 2020 (UTC)

Agree. I might add that Gene Ontology biological processes in Wikidata are a more complete and better maintained set of pathways. I would consider them the canonical database at the moment. --SCIdude (talk) 14:32, 13 February 2020 (UTC)

Why not? Indeed a pathway is an abstraction of a process, but so is a gene, which is an abstract annotation of a position on the genome. So as such it makes sense to include that abstract as a layer in the abstraction that is labeled as a pathway. So in that sense a pathway does have authors. --Andrawaag (talk) 15:18, 13 February 2020 (UTC)

I forgot that reactions are the smallest units of pathway, and that they have compounds as parts. --SCIdude (talk) 07:58, 25 December 2020 (UTC)

@GZWDer:, the alternative is to have two pages for each process, one for the pathway description (or model) about it, and one for the concept. Ontologically, that has my preference too. However, it also doubles the number of Wikidata pages, and given the current scalability issues, I am not sure what is wise here, even if ontologically better. (A good example is how we do this conflating for chemicals, where we do not distinguish the real world chemical (with physchem properties) from the chemical graph (with SMILES, InChI, etc)). This balance we find all the time in Wikidata. Is there a specific reason, besides ontological, why you like to see the split? @SCIdude:, the GO is nice for genes, but would not help as well for other types of processes. The scope is still broad, but certainly narrower. That GO is better and more complete than WikiPathways and Reactome is a bit exaggerated, but love to read the articles that showed that. But even more interesting, the GO comment is not relevant to whether the single Wikidata page per pathway/process or two: a GO pathways also has curators/authors and also cites literature. That it, it has the exact same ontological issues as brought up as GZWDer. --Egon Willighagen (talk) 16:06, 13 February 2020 (UTC)

Isn't it just a matter of an article import? I'm a Reactome pathway author and, although these works (e.g. doi:10.3180/R-HSA-9635486.1) are not imported yet in Wikidata, they might be at some time. Then link with the pathway. --SCIdude (talk) 16:36, 13 February 2020 (UTC)

Oh, I'm not talking about the journal articles. I didn't think GZWDer was talking about that. The articles cited by the WikiPathways (that are in Wikidata) are linked with a "cites", AFAIK. --Egon Willighagen (talk) 16:44, 13 February 2020 (UTC)

I'm not talking about (cited) journal articles either. Reactome has associated a DOI with each of their pathways. They are practically journal articles, with Reactome being the peer-reviewed journal, coordinating authors, their work, and their reviewers. Link to PDF associated with the pathway where I gave the DOI above: https://reactome.org/ContentService/exporter/document/event/R-HSA-9635486.pdf. Whenever that gets imported it just needs a main subject (P921) linking to the pathway item. --SCIdude (talk) 07:45, 14 February 2020 (UTC)

I don't follow, the pathway model as it currently is in Wikidata has been composed in close collaboration with the Reactome team. The model used basically aligns Wikipathways and Reactome. See this query for the reactome pathways currently covered in Wikidata and the respective models for Reactome and Wikipathways. If there are changes needed, I would suggest to discuss that on the discussion pages of those models, to also record and document those changes. --Andrawaag (talk) 08:52, 14 February 2020 (UTC)

Oh, apologies. I do like the idea of seeing databases (etc) as peer review platforms. Now, I really fully understand where you are coming from, but I honestly do not know if Wikidata is ready to splitting database records from the things the records are about. Such a split is in fact quite rare in Wikidata and am not sure it would go down well (given the scalability issues). But I already said that. And this bot page is not the right platform for a high level discussion like that. --Egon Willighagen (talk) 08:18, 15 February 2020 (UTC)

@GZWDer, AlexanderPico, Andrawaag, Egon Willighagen, Gstupp: Hello all, I just found out about this discussion after I had the same idea at Talk:Q30102291 and bothered Egon for at least a month there with my annoying pings :-). Anyway, I really do think that the "WikiPathway entries" can be separate from the molecular pathway they are describing. Therefore, I think we should move author (P50), author name string (P2093), title (P1476), and main subject (P921) to a new item, or remove them from Wikidata until someone decides to import these WikiPathway entries as "work items". Ideally,with DOIs, if they really exist for each WikiPathway entry,which would be wonderful. Vojtěch Dostál (talk) 14:46, 8 August 2022 (UTC)

The job is running right now, will be  Done in an hour or so. I eventually decided on plain removal rather than creation of new entries, because there's not a structural need for those items right now and the DOIs are not available. Vojtěch Dostál (talk) 09:39, 30 August 2022 (UTC)

problems with Reactome reactions

Hi, please have a look at the example Q101210498. If you look at the history you added Q88174316 to P527 (has part) where in Reactome one of the 1ab fragments appears. So your mapping of fragments is buggy/nonexistent. Please fix. --SCIdude (talk) 08:05, 25 December 2020 (UTC)

Some ontological relations seem to be mixed up

Hello!

First of all, thanks for the bot!

I have noticed that some entities edited by the bot could be improved in terms of modelling.

For example, Beta oxidation of octanoyl-CoA to hexanoyl-CoA (Q45317256) is listed as instance of (P31) of fatty acid beta-oxidation (Q14864021), but it really should be a subclass of (P279).

There are a few other examples where instance of (P31) is being used as a placeholder for other relations (e.g. HCMV Early Events (Q100235160) and NFG and proNGF binds to p75NTR (Q45318012).

See this SPARQL query for the P31 values used; most of them are not supposed to be there: https://w.wiki/5C4j

I'd be happy to help fixing that, just let me know and we can talk.

Best, TiagoLubiana (talk) 15:11, 22 May 2022 (UTC)