Clinical Annotations & allele information

Information from APIs

PharmGKB

• fetch clinical annotation e.g. by related chemical or gene variant symbol → array of data objects
• phenotype category in data.type
• type is a list of categories in {'Dosage', 'Efficacy', 'Metabolism/PK', 'Toxicity', 'Other'}
• implied (by us) meaning of categories
  • Metabolism/PK: info about change in metabolism, no concrete clinical action / conclusion mentioned
  • Dosage: suggestion about relative dosing and reasoning
  • Efficacy: efficacy-related clinical outcomes i.e. increase / decrease
  • Toxicity: analogous to efficacy
• data.relatedDiseases? may describe possible adverse events / symptoms (listed on PharmGKB-website as 'Phenotypes', see example)
• data.allelePhenotypes is array of objects including allele (e.g. *1), limitedEvidence (bool) and phenotype (elaborate medical description not suitable for patients)
• data.score is score as described below
  • data.scoreDetails lists individual contributing variant annotations and their scores and IDs
  • IDs can be used to query variant annotation and retrieve study parameters (e.g. p-value, number of cases / controls, biographical groups, study types, references, etc)

CPIC

• look up star allele / haplotype / diplotype function (e.g. no | decreased | normal | increased function, normal | ultrarapid | ... metabolizer)

Resources

• PharmGKB clinical annotation curation process
  • Variant annotation is a single finding from a single paper; variant annotations include study data
  • Clinical annotation is a summary of all curated information on a single variant-drug pair
• PharmGKB annotation scoring process
  • categories are added by curators

Other findings / ideas

• data.allelePhenotypes.phenotype descriptions within one clinical annotation are largely equal
  • NLP could focus on difference between star allele annotations
• combine phenotype categories (PharmGKB) and allele function categories (CPIC) for risk evaluation