A novel deep sequential learning architecture for drug drug interaction prediction using DDINet.

Area of Science:

• Computational chemistry
• Pharmacology
• Artificial intelligence in medicine

Background:

• Drug drug interactions (DDIs) pose significant healthcare challenges, leading to adverse effects and reduced treatment effectiveness.
• Predicting DDIs is crucial for patient safety and optimizing therapeutic outcomes.

Purpose of the Study:

• To introduce DDINet, a novel deep sequential learning architecture for predicting and classifying drug drug interactions (DDIs).
• To identify the underlying mechanisms of DDIs, including excretion, absorption, and metabolism.

Main Methods:

• Utilized chemical features (e.g., Hall Smart, Amino Acid count) and biochemical features extracted using the Rcpi toolkit from Simplified Molecular-Input Line-Entry System (SMILES) data.
• Employed deep sequential learning architectures, including Long Short-Term Memory (LSTM) and gated recurrent unit (GRU), with an attention mechanism.
• Trained and evaluated the DDINet model on publicly available DDI datasets from DrugBank and Kaggle.

Main Results:

• DDINet achieved high accuracy in predicting and classifying DDIs, outperforming eight existing methods.
• The model's performance was statistically validated using Confidence Interval tests and paired t-tests.
• Demonstrated the model's effectiveness in understanding DDI mechanisms.

Conclusions:

• DDINet offers a powerful computational tool for predicting and classifying drug drug interactions based on their mechanisms.
• This approach can significantly reduce the cost and time associated with traditional wet lab experiments for DDI identification.