DTIP-WINDGRU a novel drug-target interaction prediction with wind-enhanced gated recurrent unit

Area of Science:

• Computational intelligence
• Machine learning
• Drug discovery

Background:

• Drug-target interaction (DTI) identification is crucial for drug discovery but is time-consuming via lab tests.
• Computational intelligence (CI) offers essential automated prediction tools.
• Challenges include limited data, lack of negative samples, and complex existing models.

Purpose of the Study:

• To develop a novel DTI prediction model, DTIP-WINDGRU (Drug-Target Interaction Prediction with Wind-Enhanced GRU).
• To accurately determine DTIs in both labeled and unlabeled samples, surpassing traditional methods.
• To leverage advanced machine learning (ML) and deep learning (DL) for effective DTI prediction.

Main Methods:

• The DTIP-WINDGRU model incorporates pre-processing and class labeling.
• Drug-to-drug (D-D) and target-to-target (T-T) interactions initialize GRU model weights.
• Wind Driven Optimization (WDO) algorithm is used for optimal hyperparameter selection in the GRU model.

Main Results:

• The DTIP-WINDGRU model demonstrated effective DTI prediction capabilities.
• The model accurately identified interactions in both labeled and unlabeled datasets.
• Comparative analysis confirmed superior performance over existing DTI prediction techniques.

Conclusions:

• The DTIP-WINDGRU model offers a robust solution for DTI prediction.
• Extensive experimentation validated the model's efficacy across four datasets.
• The study highlights DTIP-WINDGRU's superior performance compared to current methods.