EFMSDTI: Drug-target interaction prediction based on an efficient fusion of multi-source data

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Accurate identification of Drug Target Interactions (DTIs) is crucial for understanding drug mechanisms and discovering new therapeutics.
• Computational methods integrating multi-source drug and target data accelerate drug development but often neglect individual data source contributions.
• Optimizing the fusion of multi-source data is key to enhancing DTI prediction accuracy.

Purpose of the Study:

• To propose an effective fusion strategy for multi-source drug and target data to improve Drug Target Interaction (DTI) prediction.
• To develop a computational approach, EFMSDTI, that accounts for the contribution of different data sources in DTI prediction.

Main Methods:

• Constructed 15 similarity networks using topological and semantic graphs from multi-source drug and target information.
• Fused these networks using selective and entropy weighting based on Similarity Network Fusion (SNF), prioritizing contributions to DTI prediction.
• Employed deep neural networks for low-dimensional vector embedding of drugs and targets, followed by LightGBM (Gradient Boosting Decision Tree) for final DTI prediction.

Main Results:

• EFMSDTI demonstrated superior performance with AUROC and AUPR values of 0.982, outperforming existing state-of-the-art algorithms.
• Validation of the top 1000 predicted DTIs showed a high confirmation rate, with 990 being validated.
• The approach effectively leverages the contribution of diverse data sources for accurate DTI prediction.

Conclusions:

• EFMSDTI offers a robust and accurate method for predicting Drug Target Interactions by intelligently fusing multi-source data.
• The proposed selective and entropy weighting fusion strategy effectively utilizes the contribution of different data sources, leading to improved predictive performance.
• The high validation rate of predicted DTIs underscores the practical utility of EFMSDTI in drug discovery and development.