DDI

DDI predicts and classifies drug–drug interactions to assess their clinical severity and support evaluation of interaction risk for drug safety and efficacy.

Key Features:

• OpeRational ClassificAtion (ORCA): Categorizes DDIs into five classes: contraindicated (class 1), provisionally contraindicated (class 2), conditional (class 3), minimal risk (class 4), and no interaction (class 5).
• PASS program algorithm: Uses the PASS algorithm to predict DDIs specifically for classes 1 through 3.
• PoSMNA descriptors: Employs PoSMNA (Pairs of Substances Multilevel Neighbourhoods of Atoms) chemical descriptors to represent pairs of drug substances for prediction.
• Pairwise representation: Represents drug pairs rather than single molecules to enable more comprehensive and accurate interaction predictions compared to single-molecule analyses.
• Predictive performance: Reports an average prediction accuracy of approximately 0.84 across the relevant DDI classes.
• Training data: Trained on a dataset comprising several thousand drug pairs.

Scientific Applications:

• Clinical risk assessment: Assess the clinical significance of potential DDIs between drug pairs to inform risk evaluation.
• Drug safety and efficacy evaluation: Support assessments of how interactions may affect drug metabolism and efficacy.
• Polypharmacy research: Inform research into polypharmacy and strategies to mitigate interaction risks.

Methodology:

Prediction uses the PASS program algorithm applied to PoSMNA descriptors representing drug pairs, with outputs classified according to the ORCA five-class scheme; the model was trained on several thousand drug pairs.