IDOBRU accidentally injected into IDO?

[cmungall]

http://www.ontobee.org/ontology/IDO?iri=http://purl.obolibrary.org/obo/IDO_0100199

this ID doesn't exist in IDO

the page is confusing, am I looking at IDO or IDOBRU?

[yongqunh]

I think this is because of an agreement made in early days of IDOBRU development. See: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3217896/ Copied: "As a principal in OBO foundry ontologies, an identifier is always bipartite, in the form of ID-space:Local-ID, for example: IDO:0000001. Instead of using different ID-spaces and Local-IDs for individual IDO extension ontologies, every extension ontology uses the same ID-space: "IDO", as the IDO-core. For example, an ID used for an IDOBRU term is IDO:0100246, instead of IDOBRU:0000246." As I recall, all IDOBRU IDs use the IDO_01xxxxx format.

This is some trial for us. However, in the end the ID naming strategy appears not working well ...

[linikujp]

Why the ID naming strategy is not going well? I think it is fine, if people know about the history.

On Wed, Jun 9, 2021, 00:46 Yongqun Oliver He @.***> wrote:

I think this is because of an agreement made in early days of IDOBRU development. See: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3217896/ Copied: "As a principal in OBO foundry ontologies, an identifier is always bipartite, in the form of ID-space:Local-ID, for example: IDO:0000001. Instead of using different ID-spaces and Local-IDs for individual IDO extension ontologies, every extension ontology uses the same ID-space: "IDO", as the IDO-core. For example, an ID used for an IDOBRU term is IDO:0100246, instead of IDOBRU:0000246." As I recall, all IDOBRU IDs use the IDO_01xxxxx format.

This is some trial for us. However, in the end the ID naming strategy appears not working well ...

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/OntoZoo/ontobee/issues/159#issuecomment-857368731, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACCPEPLE5KTWWKGNL2XZGI3TR3W2PANCNFSM46KNFMCA .

[cmungall]

Hi @yongqunh

Thanks for pointing me at that paper. I see indeed that was what was said in the paper. I think this was a bad decision, and the implications have not been thought through or documented.

1. There is nothing on http://obofoundry.org/ontology/ido to indicate that brucellosis terms are in there. It mentions ido core, but this isn't listed as a separate product
2. The ido OWL file does not contain IDOBRU or brucellosis terms
   • $ curl -L -s $OBO/ido.owl | grep -i brucellosis | wc -l => 0
   • you must have some special purpose procedure in ontobee to inject the idobru terms?
3. The agreement that IDOBRU IDs use the IDO_01xxxxx format is not formally encoded anywhere. What is to stop the IDO editors accidentally assigning this ID range?
4. The brucellosis terms are not present in IDO when searching ontology browsers other than ontobee.

E.g.

however there are plenty of other brucellosis terms outside idobru/ido

My recommendation is that either

1. add the terms into IDO proper
   • ido can still have a release product ido-core.owl that is upper level
   • ido can also release subsets, such as obo/ido/subsets/idobru.owl
2. OR obsolete the IDO_01xx IDs (I think the obsoletion records belong in IDO proper), make a new ontology IDOBRU, inherit from IDO, and submit IDOBRU to OBO
   • I think it is valid to submit a pre-application request to OBO before doing this

[yongqunh]

Thanks @cmungall for your checking, insights, and recommendations!

We will think these two options carefully. For now I prefer the second one to make it grow relatively independently. IDOBRU will inherit from IDO. Meanwhile, it has its own focuses and applications. When you go to that depth of representation and applications, you will find IDOBRU will not only inherit IDO, it will also inherit many terms from other ontologies such as OBI (since we want to investigate the disease in more detail), VO (we pay attention to brucellosis ontology), Protein Ontology (we care about Brucella proteins), Cell Ontology, etc.

IDOBRU is a very good representative bacterial IDO as I view it. I have done wet-lab and bioinformatics brucellosis research for over 20 years, including my PhD research, so I know the bacterium very well, and it has been my model bacterial pathogen for my ontology and bioinformatics research for a long time.

[yongqunh]

Meanwhile, I would like to acknowledge Asiyah's contriubution to the IDOBRU development. We worked together on the project for a couple of years. Thanks.

[linikujp]

Hi Chris, While what you said makes sense, we do need to carefully think about the solutions. I'd like to add @John Beverley @.> and @Shane Babcock @.> for their opinions, as they are working on coordinate and review IDO-core and extensions https://github.com/infectious-disease-ontology-extensions We will work together and find a solution that fits the most. Thanks, Asiyah

On Thu, Jun 10, 2021 at 4:48 PM Yongqun Oliver He @.***> wrote:

Meanwhile, I would like to acknowledge Asiyah's contriubution to the IDOBRU development. We worked together on the project for a couple of years. Thanks.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/OntoZoo/ontobee/issues/159#issuecomment-859052757, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACCPEPNBLX7HMS3EQCWNUJDTSEQKDANCNFSM46KNFMCA .

[linikujp]

Another note, @cmungall Most of the brucellosis terms you found are the synonyms or alternative terms of http://purl.obolibrary.org/obo/IDO_0100001. The Brucellosis AE is not Brucellosis. We can easily add the alternative terms, or mapping to other ontology terms from DOID, MONDO, EFO, Orphanet, etc. It is ironic that there are so many terms exactly the same but assigned with different IDs. Can you provide some solutions for resolving this?

[cmungall]

I don't have any magic solutions, but we can start by looking at shadow classes. @yongqunh you are listed as the contact for OAE. What is the use case for having duplicates of all diseases with "AE" appended to the end? Can you not use the core concept and use some other mechanism to indicate it's an AE? If you do think there is truly a need for the shadow concept, what is the strategy for synchronizing it with the core concept? Will you use dosdp/robot + rector normalization strategies? Currently they appear unaxiomatized

[linikujp]

AEs will be another topic. There are some overlaps between a disease and AEs, but there are a lot of things that is AE specific. We may need to gather DO, HPO, OAE and MedDRA to discuss a coordinated effort.

On Thu, Jun 10, 2021, 18:52 Chris Mungall @.***> wrote:

I don't have any magic solutions, but we can start by looking at shadow classes. @yongqunh https://github.com/yongqunh you are listed as the contact for OAE. What is the use case for having duplicates of all diseases with "AE" appended to the end? Can you not use the core concept and use some other mechanism to indicate it's an AE? If you do think there is truly a need for the shadow concept, what is the strategy for synchronizing it with the core concept? Will you use dosdp/robot + rector normalization strategies? Currently they appear unaxiomatized

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/OntoZoo/ontobee/issues/159#issuecomment-859135846, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACCPEPMZW4DOUVYLHIA2RFTTSE63RANCNFSM46KNFMCA .

[yongqunh]

That's a good question. OAE is developed by mapping to MedDRA and aims to address issues found in the MedDRA system. In the clinical setting, there are different systems including MedDRA and ICD. MedDRA focuses on adverse event, and ICD focuses on diseases. OAE terms were all manually generated, it is not developed using axioms.
I guess that it is possible to map OAE to DOID, MonDO or HPO with some design pattern and scripts. However, as Asiyah said, diseases and AEs are overlapped but differ in different ways. We may need to think about the differences carefully. MedDRA and ICD appear to focus on clinical AE and disease case report and classification. Given the focus difference, the rules of classifying them may also differ. More careful review and discussion are worthwhile.

[yongqunh]

One difference is that MedDRA uses a classification system called PT - Preferred Term. This is critical and commonly used system for clinical AE case reporting. Such a system is not used in DOID, MonDO, and HPO because the case reporting is not their purpose. Similar thing happens in ICD because it's used for clinical disease classification. The different purposes result in differnt system designs, which needs to be carefully studied. I do believe that in the end different systems should be somehow synchronized or crosslinked or axiomized using some magic protocal or strategy.

[linikujp]

PT and below are all synonyms. The MedDRA has higher terms to organize PTs: HLT, HLGT and SOC. When I was at FDA, I had training of using MedDRA and I had a design for making MedDRA RDF. Mapping to other ontologies is helpful for MedDRA, and I do have a connection with MedDRA. Let me know if you all want to have a call to make this happen. Please check https://www.slideshare.net/somucology/med-dra-basics