SMGCN: Multiple Similarity and Multiple Kernel Fusion Based Graph Convolutional Neural Network for Drug-Target...

Area of Science:

• Bioinformatics
• Computational Biology
• Drug Discovery

Background:

• Accurate identification of drug-target interactions (DTIs) is crucial for accelerating drug discovery.
• Despite extensive research, predicting DTIs remains a complex challenge.
• Existing methods often struggle to capture intricate network structures and relationships.

Purpose of the Study:

• To propose a novel computational method, SMGCN, for enhanced prediction of DTIs.
• To leverage Graph Convolutional Networks (GCN) combined with multiple similarity and multiple kernel learning.
• To improve the accuracy and efficiency of identifying potential drug-target relationships.

Main Methods:

• Developed SMGCN, integrating multiple similarity measures (Random Walk with Restart, cosine similarity) and Multiple Kernel Learning (MKL) with GCN.
• Employed GCN for extracting multi-layer, low-dimensional embedding features from network structures.
• Utilized Dual Laplace Regularized Least Squares for predicting novel DTIs via combinatorial kernels.

Main Results:

• SMGCN demonstrated significant improvements in predicting DTIs on a golden standard dataset.
• The model achieved higher performance metrics, including Area Under the Curve (AUC) and Area Under the Precision-Recall Curve (AUPR).
• The proposed method successfully identified novel DTIs, validated against the KEGG BRITE Database and scientific literature.

Conclusions:

• SMGCN offers a powerful and effective approach for predicting drug-target interactions.
• The integration of multiple similarity and MKL enhances the capability of GCN in DTI prediction.
• This method holds promise for accelerating the drug discovery pipeline by identifying novel therapeutic targets.