Predicting synthetic lethal interactions in human cancers using graph regularized self-representative matrix...

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Synthetic lethality (SL) is crucial for identifying novel cancer drug targets.
• Experimental identification of SL interactions faces significant challenges.
• Computational methods are needed to complement biological experiments for SL interaction discovery.

Purpose of the Study:

• To propose a novel computational algorithm for predicting synthetic lethal interactions.
• To enhance the accuracy of synthetic lethal interaction prediction using existing data and functional gene similarities.

Main Methods:

• Developed a graph regularized self-representative matrix factorization (GRSMF) algorithm.
• Utilized known SL interactions and Gene Ontology (GO) functional similarities among genes.
• Employed self-representation learning and matrix factorization techniques.

Main Results:

• GRSMF demonstrated superior performance in predicting potential SL interactions compared to existing methods.
• Experiments on SynLethDB data validated the effectiveness of the GRSMF algorithm.
• Case studies confirmed the ability of GRSMF to identify novel SL interactions.

Conclusions:

• GRSMF accurately predicts potential SL interactions by leveraging self-representation and gene functional similarities.
• The proposed method offers a more accurate alternative to current state-of-the-art SL interaction prediction techniques.
• GRSMF enhances the learning of self-representation matrices for improved prediction accuracy.