Wikidata

Wikidata:Property proposal/positive diagnostic predictor

< Wikidata:Property proposal

positive diagnostic predictor edit

Originally proposed at Wikidata:Property proposal/Natural science

Descriptionthe presence of the genetic variant helps to diagnose the presence of disease, used as inclusion criteria
Data typeItem
Domainsequence variant (Q15304597)
ExampleABL1 BCR::ABL (Q27653851) → chronic myeloid leukemia (Q729735)
SourceCIViC, ClinVar
Planned useTo enrich wikidata with facts on sequence variants from CIViC, ClinVar and other variant resources.
Robot and gadget jobsProteinBoxBot will update wikidata from authoritative resources on sequence variants.
See alsoCIViC, HGVS Nomenclature
Motivation

We have already added data on genes, drugs and diseases. Variants are key in linking these concepts for precision medicine and foster discussion in the community. With this property in place we can model the positive diagnostic association between a variant and a disease. Together with the provenance and the evidence. This property will allow us to connect a specific variant with evidence pertaining to its clinical significance as a predictor for the positive diagnostic of a disease phenotype. Andrawaag (talk) 19:09, 11 November 2016 (UTC)[reply]

  WikiProject Molecular biology has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead.

  Notified participants of WikiProject Medicine

Discussion
  •   Support Andrew Su (talk) 01:16, 12 November 2016 (UTC)[reply]
  •   Support This property allows incorporating in Wikidata an important annotation for genetic variation in genomic and precision medicine that is not currently possible. This addition has two great added values: the expressivity for the type of biomarker the variant is, which is key in decision-making, and to clearly track the supporting evidence for each type of biomarker to evaluate the clinical actionability of the genetic variant for diagnosis. Wikidata is a perfect platform to create community discussion on the clinical value of biomarkers, specially relevant for the biomedical and clinical research community. Nuria Queralt (talk)
  •   Support Required for modeling of variant/mutation impact in any species. Sebotic (talk) 18:43, 14 November 2016 (UTC)[reply]
  •   Support. This looks like a valuable property for the modeling of diseases and treatments. YULdigitalpreservation (talk) 14:17, 15 November 2016 (UTC)[reply]
  •   Question Let me first say I am not familiar with the bioinformatics databases, but I am wondering: How would you know which genotype is the relevant? I suppose in the case with rare genetic variants it is the presence of the rare genotype which indicates a case. But what about genetic variants with genotypes that are common? Is this obvious in the databases? I see that that there is something called "Ref. Bases" and "Var. Bases" so I suppose that item has a reference? — Finn Årup Nielsen (fnielsen) (talk) 19:33, 16 November 2016 (UTC)[reply]
  •   Question With positive diagnostic predictor and negative diagnostic predictor how would you encode that there is zero diagnostic predictor? Lets say that one study show "positive diagnostic predictor" from "BCR-ABL" to chronic myeloid leukemia (Q729735) and then someone encodes that in the "BCR-ABL" item, while later another study finds "negative diagnostic prediction" for the same relation. After several other studies a meta-analysis finds "no diagnostic prediction" for the relation? You seem to have no way of expressing this? — Finn Årup Nielsen (fnielsen) (talk) 19:44, 16 November 2016 (UTC)[reply]
    • One say could be to just have "diagnostic predictor" and then encode the direction or effect size in a qualifier. — Finn Årup Nielsen (fnielsen) (talk) 19:48, 16 November 2016 (UTC)[reply]
      • @Finn Årup Nielsen (fnielsen) In that case we would add that information with statement disputed by (P1310) as a qualifier to the statement of positive/negative diagnostic predictor and link it to the PMID or URL of the study. Emitraka (talk) 21:24, 16 November 2016 (UTC)[reply]
      • A predictor can only predict in two directions, right: positive or negative. And there are only two possible values for the effectiveness as a predictor: has predictive value or has not. I think it is less verbose and clearer to see in the property if the genetic variant is a positive or a negative predictor, and to see how many sources support the predictive value of the variant and how many studies dispute it using this statement disputed by (P1310) qualifier. Also, in this way, one can have a hint about the potential clinical relevance of the variant with this evidence chain. In addition, i think it allows to query this information in an easier way than using several qualifiers what can make the querying more cumbersome imo.. NuriaQueralt (talk) 5:29, 17 November 2016 (UTC)
      • One element that is important and can't really be expressed in a single property is the relative predictive power of the sign under consideration. A predictor could vary from altering the odds of the prediction by 1% all the way to being 100% accurate by itself. I think having these fairly high level relations as proposed is useful but that when applying them we should strive to make effective use of qualifiers to add important contextual information such as the methods used to identify the predictive association (e.g. a clinical trial, or a particular kind of screen) as well as recording the estimated predictive power for the claim. I   Support this proposal but suggest that we provide a template qualifier structure to make the statements created most useful. --I9606 (talk) 20:05, 18 November 2016 (UTC)[reply]
  •   SupportPutmantime (talk) 21:57, 18 November 2016 (UTC)[reply]
  • @Andrawaag, Andrew Su, NuriaQueralt, Sebotic, YULdigitalpreservation: @Fnielsen, Emitraka, I9606, Putmantime:   Done ArthurPSmith (talk) 15:36, 21 November 2016 (UTC)[reply]
Retrieved from "https://www.wikidata.org/w/index.php?title=Wikidata:Property_proposal/positive_diagnostic_predictor&oldid=410038993"