[Paper Review] IGLOO: Integrating global and local biological network alignment

Lei Meng, Joseph Crawford+2 more|arXiv (Cornell University)|Apr 20, 2016

Bioinformatics and Genomic Networks27 references17 citations

TL;DR

IGLOO proposes a novel network alignment method that integrates local (LNA) and global (GNA) network alignment by using a high-functional-quality LNA seed to guide topological expansion via GNA, achieving a superior trade-off between functional and topological alignment quality compared to state-of-the-art LNA and GNA methods.

ABSTRACT

Analogous to genomic sequence alignment, biological network alignment (NA) aims to find regions of similarities between molecular networks (rather than sequences) of different species. NA can be either local (LNA) or global (GNA). LNA aims to identify highly conserved common subnetworks, which are typically small, while GNA aims to identify large common subnetworks, which are typically suboptimally conserved. We recently showed that LNA and GNA yield complementary results: LNA has high functional but low topological alignment quality, while GNA has high topological but low functional alignment quality. Thus, we propose IGLOO, a new approach that integrates GNA and LNA in hope to reconcile the two. We evaluate IGLOO against state-of-the-art LNA (NetworkBLAST, NetAligner, AlignNemo, and AlignMCL) and GNA (GHOST, NETAL, GEDEVO, MAGNA++, WAVE, and L-GRAAL) methods. We show that IGLOO allows for a trade-off between topological and functional alignment quality better than the existing LNA and GNA methods considered in our study.

Motivation & Objective

• Address the complementary nature of local (LNA) and global (GNA) network alignment, where LNA excels in functional quality but lacks topological conservation, while GNA achieves high topological alignment at the cost of functional quality.
• Overcome the inherent trade-off between topological and functional alignment quality by combining the strengths of both LNA and GNA approaches.
• Develop a generalizable framework that can integrate any existing LNA and GNA methods to improve alignment quality without requiring algorithmic redesign.
• Demonstrate that the integration leads to a more balanced and improved trade-off between functional and topological alignment quality than standalone LNA or GNA methods.

Proposed method

• IGLOO uses a high-quality local alignment (generated by an existing LNA method) as a seed alignment to initiate the alignment process.
• The seed alignment is expanded using a global network alignment (GNA) algorithm to improve topological conservation while preserving functional similarity.
• The method supports many-to-many node mappings (like LNA) to maintain functional sensitivity, while enabling large-scale conserved subgraph identification (like GNA).
• Node similarity scores are computed via a node cost function (NCF), which incorporates both topological and sequence-based information.
• IGLOO is modular and generalizable, allowing integration of any existing LNA and GNA methods as components.
• The algorithm is evaluated using a combination of topological (edge conservation, node coverage) and functional (F-score of protein function prediction) metrics.

Experimental results

Research questions

• RQ1 Can integrating LNA and GNA methods yield a network alignment with improved balance between functional and topological quality compared to standalone LNA or GNA?
• RQ2 Does using a high-functional-quality LNA seed improve the functional alignment quality of a GNA-based expansion process without degrading topological quality?
• RQ3 How does IGLOO compare to state-of-the-art LNA and GNA methods in terms of both functional and topological alignment quality across diverse PPI networks?
• RQ4 What is the computational overhead of IGLOO relative to existing LNA and GNA methods, and does it scale reasonably with network size and complexity?
• RQ5 Can IGLOO consistently outperform individual LNA and GNA methods across multiple network types and confidence levels?

Key findings

• IGLOO 4 achieved the highest functional alignment quality among all tested methods, significantly outperforming all considered LNA and GNA methods.
• IGLOO 0 demonstrated statistically significant superiority in topological alignment quality over four GNA methods (GHOST, NETAL, GEDEVO, MAGNA++), and was the strongest GNA-comparable performer.
• IGLOO 4 was statistically significantly superior to all four tested GNA methods in functional alignment quality, and to all but one (GEDEVO) in topological alignment quality.
• IGLOO 0 was the only version to achieve statistically significant topological superiority over NETAL and MAGNA++, which are among the top-performing GNA methods.
• IGLOO's running time was comparable to or faster than most LNA and GNA methods, with only slight overhead due to integration with AlignMCL and NETAL.
• The method demonstrated robustness across four PPI network datasets of varying interaction types and confidence levels, confirming its generalizability and effectiveness.