Nearest Neighbor Networks: clustering expression data based on gene neighborhoods

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Microarray data analysis presents challenges in translating large-scale gene expression data into biological insights.
• Gene clustering is a crucial first step, but traditional methods like hierarchical and K-means clustering have limitations.
• Existing methods may identify spurious clusters or be biased towards a narrow range of biological functions.

Purpose of the Study:

• To develop a novel algorithm for clustering genes with similar expression profiles from microarray data.
• To address the limitations of existing gene clustering techniques in identifying functionally coherent gene sets.
• To improve the precision and breadth of biological processes captured in gene clusters.

Main Methods:

• Developed Nearest Neighbor Networks (NNN), a graph-based clustering algorithm.
• NNN utilizes overlapping cliques within an interaction network derived from mutual nearest neighborhoods.
• The algorithm focuses on connectivity and mutual similarity to form gene clusters.

Main Results:

• NNN successfully generated functionally coherent gene clusters with high precision.
• The clusters identified by NNN represented a broader selection of biological processes compared to eight other methods.
• The method effectively captures highly connected gene clusters even when spatially separated in expression space.

Conclusions:

• Nearest Neighbor Networks (NNN) is an effective and valuable algorithm for grouping functionally related genes.
• The algorithm's simplicity and success in analyzing large datasets make it attractive for gene expression studies.
• NNN demonstrates a high precision in identifying functionally related genes across a wide spectrum of biological functions.