Multi-Neighborhood Learning for Global Alignment in Biological Networks.

Area of Science:

• Bioinformatics
• Computational Biology
• Network Science

Background:

• Global alignment of biological networks (GABN) is crucial for understanding species evolution, orthology detection, and genetic analyses.
• Existing GABN methods struggle to balance conservation of biological and topological network structures.

Purpose of the Study:

• To propose a novel multi-neighborhood learning method (CLMNA) for GABN.
• To address the challenge of achieving a good tradeoff between biological and topological structure conservation in network alignment.

Main Methods:

• Modeled GABN as an optimization problem based on weighted similarity.
• Integrated first-proximity, second-proximity, and individual-aware proximity learning algorithms.
• Evaluated conserved biological and topological similarities.

Main Results:

• CLMNA demonstrated superiority over state-of-the-art network alignment algorithms in systematic experiments.
• Experiments were conducted on 10 pairs of biological networks across 5 species.
• CLMNA effectively improved the performance of compared algorithms as a refinement method.

Conclusions:

• CLMNA offers a superior approach to GABN by effectively balancing structural conservation.
• The proposed method advances the field of biological network analysis and its applications.