NeverShow: Metric treats Multiple Sclerosis Automated Failure

[cbizon]

NeverShow: Metric treats Multiple Sclerosis

This answer is coming up for several ARAs because RTX-KG2 is returning an edge:

{
    "subject": "UMLS:C0699680",
    "object": "MONDO:0005301",
    "predicate": "biolink:treats",
    "sources": [
        {
            "resource_id": "infores:rtx-kg2",
            "resource_role": "aggregator_knowledge_source",
            "upstream_resource_ids": [
                "infores:semmeddb"
            ]
        },

Based on these kinds of sentences: "Correlation of diffusion tensor and dynamic perfusion MR imaging metrics in normal-appearing corpus callosum: support for primary hypoperfusion in multiple sclerosis."

Here "Metric" is being misinterpreted as an antifungal that I guess is called "Metric"

[saramsey]

Thanks, we will take a look at this.

[dkoslicki]

It appears that this is an underlying SemMedDB issue, as it is confusing metric (the drug) with metric (the measurement). Ongoing efforts to clean up SemMedDB are progressing, but at this moment this isn't an ARAX or RTX-KG2 problem, but an underlying KS issue

[sierra-moxon]

@andrewsu - would this ("metric") be a good candidate for a blocklist?

[andrewsu]

skimming usages of UMLS:C0699680 "Metric" in https://biothings.transltr.io/semmeddb/query?q=subject.umls:C0699680%20OR%20object.umls:C0699680, I can't find an example where this is not an NER error. So yes, I think this would be a good blocklist candidate. (And I'm assuming you meant a global blocklist, not a semmeddb-specific one...)