INDI: a computational framework for inferring drug interactions and their associated recommendations

Area of Science:

• Pharmacology and Bioinformatics
• Drug Discovery and Development

Background:

• Drug-drug interactions (DDIs) are critical in drug development and administration.
• Current computational methods primarily focus on pharmacokinetic DDIs mediated by Cytochrome P450 (CYP) enzymes.
• There is a need for methods that can infer both pharmacokinetic and pharmacodynamic DDIs.

Purpose of the Study:

• To introduce INDI, a novel computational method for inferring both pharmacokinetic and pharmacodynamic DDIs.
• To evaluate INDI's performance in predicting DDIs and their severity.
• To assess the prevalence of known and predicted DDIs in clinical settings.

Main Methods:

• Development of INDI, a prediction method for inferring CYP-related and non-CYP drug interactions.
• Cross-validation using AUC (area under the receiver-operating characteristic curve) to assess prediction accuracy (≥0.93).
• Application of INDI to the FDA adverse event reporting system and analysis of hospitalized patients' medication data.

Main Results:

• INDI achieves high specificity and sensitivity in DDI prediction (AUC ≥0.93).
• 53% of drug events in the FDA system were linked to known (41%) or predicted (12%) DDIs.
• 18% of hospitalized patients received known or predicted severely interacting drugs, correlating with higher hospitalization frequency.

Conclusions:

• INDI effectively predicts a wide range of DDIs, including pharmacokinetic and pharmacodynamic types.
• A significant proportion of clinical drug events and patient medications involve potentially severe DDIs.
• INDI provides a valuable tool for identifying and exploring DDIs, supporting clinical practice and pharmaceutical research.