treatment edges in CIDO

[cmungall]

I'm looking at treatment-for edges in CIDO: http://purl.obolibrary.org/obo/CIDO_0000012

there are two edges:

• COVID-19 drug subClassOf : treatment for some COVID-19 disease process
• Hydroxychloroquine Oral Tablet subClassOf : treatment for some COVID-19 disease process

there are no actual drugs under 'COVID-19 drug':

So the only COVID treatment in CIDO is HCQ

Now I know that OWL follows the open world assumption, absence of a statement does not mean false. However, I think it's really important that as the sole COVID-19 specific ontology in OBO that CIDO makes best effort attempts at completeness.

The choice of HCQ as the single drug is problematic, it's controversial, not supported by FDA, etc.

I think there should at least be more metadata and provenance - how is the treatment-for edges curated? What is the evidence for each assertion? What does it mean to be a treatment for anyway? What is the success rate? Is this approved treatment or any treatment? Is bleach a treatment, because someone tried it?

Note that dedicated compound resources such as pubchem curate lists of relevant compounds with clear criteria, e.g compounds currently used in clinical trials:

https://pubchemdocs.ncbi.nlm.nih.gov/covid-19

Remdesivir is top of the list:
https://pubchem.ncbi.nlm.nih.gov/#tab=compound&query=covid-19%20clinicaltrials

This is much more useful

I think at the least CIDO should recommend users to go to a resource such as pubchem, as currently users may be confused or worse

[cmungall]

A minor aside is that treatment-for does not have a domain specified, and the domain is currently a mixture of roles and physical entities. This is not the main issue here, but I thought I would mention

[cmungall]

Another aside:

HCQ oral tablet: http://www.ontobee.org/ontology/CIDO?iri=http://purl.obolibrary.org/obo/DRON_00020600

this says it is an anti-malarial which is correct, but does not list other uses. It lacks information HCQ is used to treat SLE which is a well-known use. Again, by the open world assumption this is fine, but I think there is a commitment to providing complete information within reason.

Drugbank has a more complete list of diseases treated, including SLE
https://www.drugbank.ca/drugs/DB01611

(drugbank also includes information that HCQ is undergoing CTs for COVID)

Similarly, CIDO only lists one adverse effect (arhythmia). Under Open World Assumption this is not incorrect, but again it behooves us to include complete information where this is possible

DrugCentral provides more comprehensive informaation, it lists arrhythmia and many others, plus quantitative information.

I get that CIDO provides value in integrating this information in one place, but when it is incomplete to the point of being confusing or misleading it is difficult to justify. I think it may be a good idea to be more upfront about the content of CIDO, current curation practices, and expectations of completeness.
http://drugcentral.org/drugcard/1395

[wdduncan]

@cmungall Thanks for pointing out the need for more information.
Going back to your original point (I think) about adding metadata to the "edge" treatment for some COVID-19 disease process.
How would you propose we do this? Reification (rdf or owl) is one option.

[yongqunh]

@cmungall Thanks for pointing out! We are currently working hard on these. We have manually annotated over 100 chemical drugs that have been experimentally or clinically proven effective against human coronaviruses including SARS-CoV, SARS-CoV-2, and MERS-CoV. These drugs have been moved to ChEBI, DrON, and NDF-RT if possible. We are now adding more annotations and axioms to these drugs. A lot of work to do. But we are progressing quite fast.
More suggestions and comments are welcome and appreciated!

[yongqunh]

@cmungall
We need to generate consensus ontology design patterns and use them to represent and incorporate COVID-19 related data from different resources including:
https://pubchemdocs.ncbi.nlm.nih.gov/covid-19
https://www.drugbank.ca/covid-19
etc.
I think we can have more meetings to discuss these design patterns and ensure consensus is achieved.

[cmungall]

It's great you are manually curating these compounds. How are you coordinating with existing authoritative databases? Do you have a distribution plan beyond CIDO? Very few people will look in an ontology to see these kinds of annotations.

While working this and the design patterns out you may wish to remove your one misleading treatment link.

Note in kg-covid-19 we have python to ingest some of the aforementioned sources into a KG form, these can easily be treated as OWL axioms (but we prefer to keep this kind of information in the KG)

[yongqunh]

@cmungall thanks for your comment. we have removed the treatment link that links HCQ to COVID.
Other improvements are ongoing ....

[yongqunh]

@cmungall I realized that the CIDO 'treatment for' relation can be changed to the RO term 'is substance that treats' (RO_0002606).