Synergistic Drug Combination Prediction via Dual-Level Feature Aggregation and Knowledge Graph-Based Deep Neural...

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Identifying synergistic drug combinations is crucial for effective cancer treatment but computationally challenging.
• Existing methods often lack comprehensive data integration, overlooking complex biological interactions.
• There's a need to incorporate intrinsic drug/cell line properties and broader biological relationships for accurate synergy prediction.

Purpose of the Study:

• To develop a novel computational framework, LGSyn, for predicting synergistic drug combinations.
• To integrate diverse biological data, including local and global features, for enhanced prediction accuracy.
• To evaluate LGSyn's performance against state-of-the-art models.

Main Methods:

• LGSyn integrates local features (molecular fingerprints, descriptors, gene expression) and global features (drug-protein, protein-cell line, protein-protein, cell line-tissue interactions).
• Three fusion strategies were developed to effectively combine local and global information.
• A deep neural network was employed for training and synergy prediction using the fused features.

Main Results:

• The proposed LGSyn framework demonstrated superior accuracy and stability in predicting drug synergy.
• LGSyn outperformed current state-of-the-art computational methods.
• Experimental results validate the effectiveness of integrating multi-perspective biological data.

Conclusions:

• LGSyn offers a powerful and effective approach for predicting synergistic drug combinations by leveraging comprehensive biological knowledge.
• The integration of local and global features significantly improves the accuracy and stability of synergy prediction.
• The developed framework has the potential to advance cancer therapeutic strategies.