Translational Knowledge Discovery Between Drug Interactions and Pharmacogenetics.

Area of Science:

• Pharmacogenomics
• Computational Biology
• Drug Discovery

Background:

• Clinical application of drug-drug interaction (DDI) studies is hindered by knowledge gaps.
• Consolidating diverse DDI evidence and linking it to clinical outcomes remains challenging.

Purpose of the Study:

• To develop information retrieval (IR) and information extraction (IE) models to identify DDI and drug-gene interaction (DGI) evidence.
• To categorize DDI evidence into in vitro pharmacokinetic (PK), clinical PK, and clinical pharmacodynamic (PD) studies.
• To integrate DDI and DGI evidence to uncover novel pharmacogenetic associations.

Main Methods:

• Utilized IR models to scan 25 million PubMed abstracts for DDI and DGI evidence.
• Employed IE models to extract DDI-pairs and DGI-pairs from retrieved abstracts.
• Performed overlapping analysis to identify unique DDI-pairs across different evidence types.

Main Results:

• Identified 986 unique DDI-pairs across in vitro PK, clinical PK, and clinical PD evidence.
• Extracted 2,157 DDI-pairs and 13,012 DGI-pairs from various evidence sources.
• Discovered 119 and 18 novel pharmacogenetic hypotheses linked to CYP3A and CYP2D6, respectively, by integrating DDI and DGI data.

Conclusions:

• The developed IR and IE models effectively retrieve and categorize DDI and DGI evidence.
• Integration of DDI and DGI data facilitates the discovery of new pharmacogenetic hypotheses.
• This approach enhances understanding of DDI mechanisms and aids in clinical translation.