An Optimised Mobilenet V2 Attention Parallel Network for Predicting Drug-Drug Interactions Through Combining Local...

Area of Science:

• Pharmacology
• Artificial Intelligence
• Computational Biology

Background:

• Drug-drug interactions (DDIs) pose significant risks to patient safety and complicate drug development.
• Accurate DDI prediction is crucial for effective medication management and risk reduction.

Purpose of the Study:

• To develop and evaluate a novel technique, MV2SAPCNNO, for improving the precision of DDI prediction.
• To enhance the accuracy and efficiency of DDI prediction models.

Main Methods:

• Data preprocessing including normalization and noise reduction.
• Feature extraction using MobileNetV2 with simplicial attention network (MV2SAN) for local and global features.
• Parallel convolutional neural network (PCNN) processing optimized by the narwhal optimizer (NO) for parameter tuning and error minimization.

Main Results:

• The MV2SAPCNNO model demonstrated superior performance compared to existing DDI prediction models.
• Achieved enhanced accuracy, precision, recall, and F-score metrics.
• The narwhal optimizer improved convergence efficiency and reduced computational time.

Conclusions:

• MV2SAPCNNO offers an efficient and accurate approach to DDI prediction.
• The model's performance contributes to safer medication administration and drug development.
• Findings support enhanced patient safety in clinical practice.