An Evidence-Based Framework for Evaluating Pharmacogenomics Knowledge for Personalized Medicine.

Area of Science:

• Pharmacogenomics
• Biomedical Informatics

Background:

• Clinical annotations in the Pharmacogenomics Knowledgebase (PharmGKB) summarize variant-drug associations and are assigned a Level of Evidence (LOE).
• Maintaining consistency in LOE assignment becomes challenging as more evidence is curated by multiple curators over time.

Purpose of the Study:

• To develop and implement an automated scoring system for consistent and reproducible LOE assignment to PharmGKB clinical annotations.
• To improve the transparency and robustness of pharmacogenomic data interpretation.

Main Methods:

• A scoring system was developed to automate LOE assignment based on attributes of variant annotations (e.g., study size, P value, association findings).
• Scores from clinical guidelines and FDA-approved drug labels were also incorporated.
• Scores of all attached annotations are summed to calculate a total score, which determines the clinical annotation's LOE.

Main Results:

• The scoring system automates LOE assignment, increasing transparency, consistency, and reproducibility.
• Standardized writing of clinical annotations combined with the scoring system ensures PharmGKB remains a reliable source of pharmacogenomic information.

Conclusions:

• The automated scoring system effectively standardizes LOE assignment for PharmGKB clinical annotations.
• This system enhances the reliability and usability of pharmacogenomic data for clinical decision-making.