Predicting drug side effects by multi-label learning and ensemble learning.

Area of Science:

• Pharmacology and Bioinformatics
• Computational Drug Discovery
• Machine Learning in Medicine

Background:

• Predicting drug side effects is crucial for drug discovery and safety.
• Existing machine learning methods for side effect prediction have limitations.
• Multi-label learning and feature selection can enhance prediction accuracy and interpretability.

Purpose of the Study:

• To develop an improved method for predicting drug side effects.
• To identify critical drug-related feature dimensions associated with side effects.
• To construct high-accuracy multi-label prediction models for adverse drug reactions.

Main Methods:

• Proposed a novel feature selection-based multi-label k-nearest neighbor (FS-MLKNN) method.
• FS-MLKNN simultaneously determines critical feature dimensions and builds prediction models.
• Developed an ensemble learning model by integrating individual FS-MLKNN models for enhanced performance.

Main Results:

• FS-MLKNN achieved good performance with explainable results in computational experiments.
• The ensemble learning model integrated with FS-MLKNN demonstrated superior performance compared to state-of-the-art methods.
• The proposed methods showed effectiveness on benchmark datasets for drug side effect prediction.

Conclusions:

• FS-MLKNN and the ensemble method are promising tools for accurate drug side effect prediction.
• These methods offer insights into the causes of side effects by identifying critical feature dimensions.
• The source code and datasets are publicly available for further research.