MPASL: multi-perspective learning knowledge graph attention network for synthetic lethality prediction in human...

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Synthetic lethality (SL) is crucial for identifying anti-cancer drug targets.
• Experimental identification of SL interactions is resource-intensive and inefficient.
• Developing accurate computational prediction methods for SL is highly significant.

Purpose of the Study:

• To propose MPASL, a multi-perspective learning knowledge graph attention network.
• To enhance the prediction accuracy of synthetic lethality interactions.
• To provide an efficient computational tool for drug target discovery.

Main Methods:

• MPASL employs knowledge graph hierarchy propagation to explore gene-related nodes.
• Knowledge graph ripple propagation expands gene representations using existing SL preference sets.
• MPASL learns gene representations from gene-entity and entity-entity perspectives, refining them with discrepancy contrastive learning.

Main Results:

• MPASL demonstrates superior performance compared to existing state-of-the-art methods.
• Experimental results validate MPASL's effectiveness in predicting synthetic lethality interactions.
• Case studies confirm the method's utility in identifying gene-gene SL relationships.

Conclusions:

• MPASL offers a powerful and accurate computational approach for synthetic lethality prediction.
• The method advances the discovery of anti-cancer drug targets.
• MPASL provides a valuable tool for bioinformatics and computational drug discovery research.