CME-KGDTI: integrating clustered mutations into knowledge graph embedding for drug-target interaction prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Drug-target interaction (DTI) prediction is vital for drug discovery and precision medicine.
• A biologically enriched heterogeneous knowledge graph (KG) was constructed, integrating mutations, synthetic lethal interactions, drug structures, protein sequences, and functional annotations.
• This framework aims to identify diagnostic signatures and precision therapeutics using multi-layered biological network factors.

Purpose of the Study:

• To develop a computational model for predicting drug-target interactions (DTIs) by leveraging a biologically enriched knowledge graph.
• To enable the identification of actionable diagnostic signatures and precision therapeutic strategies.
• To create a user-friendly platform for exploring mutation detection, target prioritization, and mechanistic insights.

Main Methods:

• Constructed a heterogeneous knowledge graph (KG) integrating diverse biological data.
• Applied graph embedding techniques (TransE, RotatE, DistMult, Node2vec, R-GCN) to represent KG entities.
• Utilized deep learning models (DNN, NFM, AutoInt) with dimensionality reduction for DTI prediction, developing the CME-KGDTI model.
• Developed the CME-KGDTI platform for comprehensive analysis.

Main Results:

• The CME-KGDTI model demonstrated superior performance, especially in a challenging protein cold-start scenario.
• Multi-source biological information enhanced positive sample identification and model generalization.
• The CME-KGDTI platform integrates resources for mutation identification, genetic networks, DTI prediction, and multi-omics analysis.

Conclusions:

• The CME-KGDTI model, incorporating biological features, achieves high accuracy and robust generalization for drug target discovery.
• The CME-KGDTI platform serves as a flexible, interactive tool to advance precision oncology research.
• This approach complements existing methods in drug target identification and development.