SSELM-neg: spherical search-based extreme learning machine for drug-target interaction prediction

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in drug discovery

Background:

• Experimental drug discovery is costly and time-consuming.
• Accurate identification of drug-target interactions (DTIs) is crucial for efficient drug development.
• Existing machine learning methods for DTI prediction face challenges like class imbalance and parameter optimization.

Purpose of the Study:

• To develop an efficient and effective framework for predicting drug-target interactions (DTIs).
• To address the challenges of class imbalance and parameter optimization in DTI prediction models.

Main Methods:

• Proposed a framework named SSELM-neg for DTI prediction.
• Utilized a screening approach for selecting high-quality negative samples.
• Employed a spherical search approach for optimizing extreme learning machine (ELM) parameters.

Main Results:

• The SSELM-neg framework demonstrated superior performance in DTI prediction.
• Achieved high accuracy in 10-fold cross-validation experiments.
• Outperformed state-of-the-art methods on enzyme, G-protein coupled receptor, ion channel, and nuclear receptor datasets, with AUC scores ranging from 0.969 to 0.993 and AUPR scores from 0.946 to 0.991.

Conclusions:

• The screening approach effectively resolved the class imbalance problem by generating high-quality negative samples.
• Optimizing ELM parameters using spherical search enhanced DTI identification accuracy.
• The developed models significantly outperformed existing state-of-the-art methods for DTI prediction.