A Knowledge-Based System for Intelligent Support in Pharmacogenomics Evidence Assessment: Ontology-Driven Evidence Representation and Retrieval 2017 Joint Summits on Translational Science | AMIA Tuesday, March 28, 2017 S17: Papers – Clinical Integration ChiaJu (Cheryl) Lee, PhD Disclosure > I disclose that I have no relevant financial relationships with commercial interests. 2 Background and Significance Precision Medicine Takes Individual Variation into Account Harmful but effective Harmful Pharmacogenomics (PGx) • How genetic variants affect a person’s response to a drug • Takes into account individual variability in genes • For safe & effective medication prescribing Neither harmful nor effective Safe and effective Source: http://sites.google.com/site/genomicssok/home 3 Background and Significance Widespread Integration of PGx in Everyday Clinical Practice Is Still Lacking > Projects in support of preemptive genetic testing – – – – PREDICT1 The 1200 Patients2 PG4KDS3 RIGHT Protocol4 Insufficient evidence to recommend clinical validity and utility of a genetic testing5-8 The need for development of knowledge bases PharmGKB9, ClinVar10, ClinicalTrials.gov 4 [1]PMID: 22588608, [2]PMID: 22929923, [3]PMID: 24619595, [4]PMID: 24388019 [5]PMID: 16013993, [6]PMID: 20981006, [7]PMID: 21884816, [8]PMID: 21619429 [9]PMID: 20350130, [10]PMID: 24234437 Research Hypothesis Proposal: Knowledge-Based System (KBS) for Providing Computable PGx Evidence > With the rapid growth of genomic research, the availability of evidence may not be a major concern > Real challenge is how to make effective use of existing study results to support timely decision making > Hypothesis: a knowledge-based system to facilitate effective and efficient evidence assessment of PGx adoption 5 Methods Improving the Systematic Review Pipeline Natural language processing, machine learning, text mining11-13 Conduct a literature search Screen to identify relevant studies Extract essential data from studies Knowledge-based system (KBS) Synthesize the extracted data 6 Rate the quality and strength of evidence [11]PMID: 20920176, [12]PMID: 25005128, [13]PMID: 26073888 Interpret the results Methods Analogy between TINKERTOY® and Knowledge-Based System Constructors e.g., connector Constructs e.g., rod Source: http://www.knex.com/30-model-super-building-set The more complex construction demands more constructs and constructors 7 Methods Knowledge-Based System Is Capable of Providing Reasoning Services Constructs i.e., classes, properties, instances Knowledge Representation Language formal representation Constructors i.e., property restrictions, set operators Ontology (TBox) Asserted Individuals (ABox) reasoning Knowledge Base Knowledge-Based System Expressivity in representation vs. Efficiency in reasoning Applications 8 Retrieval Reasoner Classification Draw Inferences Methods Strategy to Develop a KBS de novo 1. Conceptual modeling – Describe a domain through a conceptual model 2. Ontology development – Convert the model to ontology constructs 3. Ontology-based knowledge base development – Encode asserted individuals by combining ontology constructs with constructors 4. Evaluation – – 9 Correctness in representation Efficiency in reasoning Methods Prototype PGx-KBS for Clopidogrel and Warfarin (2) Ontology development (1) Conceptual modeling (4) KBS evaluation (3) KB development A conceptual model composed of 9 information modules Individual P Publication 10 Individual S Study Population Study Design Drug Therapy Risk of Bias Assessment Individual E Comparison Genetic Variation Outcome Effect Estimation Results A Piece of Computable Individual Evidence 7 constructs (3 classes, 4 properties), 2 constructors (some, and), 1 data value (i.e., 28) Outcome measure was the percentage of time of international normalized ratio (INR) in the therapeutic range up to the follow-up of 28 days14. [14]PMID: 24251361 11 Results Evaluation of Prototype PGx-KBS for Retrieval Effectiveness and Efficiency 73 computable Publications 82 computable Studies 445 computable Evidence 33 defined classes 33 IC [15]PMID: 22929815, [16]PMID: 22591668, [17]PMID: 21816733 [18]PMID: 21693476, [19]PMID: 20845077, [20]PMID: 20620727 [21]PMID: 20351750, [22]PMID: 22203539, [23]PMID: 22757746 Singh15 Jang16 Bauer17 Zabalza18 # of MA included 9 6 4 4 # of E retrieved for all included MAs 57 58 44 Computing time (sec) 23 21 100% 100% SRs Precision 11 Jin19 Hulot20 Sofi21 Holmes22 Yamaguchi23 3 3 2 1 1 31 19 22 23 31 16 21 16 18 17 18 11 9 100% 100% 100% 100% 100% 100% 100% Contributions Major Contributions > Among the first to employ a KBS to improve efficiency, precision and transparency in the systematic review process – Ontology for formal representation of PGx knowledge and inclusion criteria – KB for accumulation of computable knowledge – Reasoner for efficient retrieval > Exploited the expressivity and reasoning ability of OWL 2 DL – Advanced constructors of OWL 2 DL > Benefit the evidence-based practice of genomics medicine 13 Limitations & Future Work Major Limitations and Future Work > The whole process is not seamless in practice – No efforts have been undertaken to develop plug-ins to automatically export retrieval results from Protégé to existing statistical software that supports meta-analysis. > Scope is limited to clinical validity and utility of clopidogrel and warfarin PGx – Future Work: expand from germline to somatic/tumor PGx e.g., cancer PGx for targeted therapy selection 14 Acknowledgements Acknowledgements > > > > > Peter Tarczy-Hornoch, MD, FACMI Beth Devine, Pharm.D., MBA, PhD James Brinkley, MD, PhD, FACMI John Horn, Pharm.D., FCCP University of Washington Colleagues – Precision Medicine Informatics Group > This work was supported in part by U01 HG006507, Biomedical and Health Informatics Program and Pharmaceutical Outcomes Research & Policy Program 15 Thank you! Questions? [email protected] 16
© Copyright 2026 Paperzz