Identifying "Many-to-Many" Relationships between Gene-Expression Data and Drug-Response Data via Sparse Binary...

Area of Science:

• Pharmacology
• Genomics
• Bioinformatics

Background:

• Identifying gene-drug patterns is crucial for understanding disease mechanisms and advancing drug discovery.
• High-throughput technologies generate vast pharmacological and genomic data, offering opportunities to explore gene-drug interactions.
• Previous methods often lack integration of prior biological knowledge, limiting their ability to infer complex gene-drug relationships.

Purpose of the Study:

• To develop a novel computational model for decoding gene-drug relationships from large-scale heterogeneous data.
• To integrate gene expression, drug response data, and network-based prior information for improved pattern identification.
• To validate the proposed model's effectiveness in revealing biologically relevant gene-drug patterns.

Main Methods:

• Proposed a network-guided sparse binary matching (NSBM) model.
• Jointly analyzed large-scale gene-expression and drug-response data.
• Incorporated prior gene and drug information using network-based regularization within a convex quadratic minimization framework.

Main Results:

• The NSBM model demonstrated superior performance compared to two benchmark methods in extensive experiments.
• Experiments were conducted on both synthetic and empirical datasets to validate the model's efficacy.
• Posterior validation using gene-ontology and enrichment analysis confirmed the model's ability to reveal significant gene-drug patterns.

Conclusions:

• The network-guided sparse binary matching (NSBM) model is effective for uncovering gene-drug patterns in large-scale heterogeneous data.
• Integrating network-based prior information enhances the identification of biologically relevant gene-drug relationships.
• The NSBM model offers a powerful approach for drug discovery and understanding disease mechanisms.