DeepACTION: A deep learning-based method for predicting novel drug-target interactions.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug-target interactions (DTIs) are crucial for drug development and repurposing.
• Experimental prediction of DTIs is resource-intensive.
• Computational methods offer a faster alternative for identifying DTIs.

Purpose of the Study:

• To propose deepACTION, a novel deep learning model for predicting unknown drug-target interactions.
• To address challenges in DTI prediction, including high dimensionality and class imbalance.

Main Methods:

• Representing drugs and proteins using structural and sequence descriptors.
• Employing the MMIB technique for dataset balancing.
• Utilizing a LASSO model for dimensionality reduction.
• Training a convolutional neural network (CNN) for DTI prediction.

Main Results:

• The deepACTION model achieved a high Area Under the Curve (AUC) of 0.9836 on the DrugBank dataset.
• The model demonstrated superior performance compared to existing methods via 5-fold cross-validation.
• The MMIB technique and LASSO effectively handled data imbalance and high dimensionality.

Conclusions:

• The deepACTION model accurately predicts novel drug-target interactions.
• This computational approach can significantly accelerate drug discovery and repurposing efforts.
• The model provides valuable insights for future drug development initiatives.