Genome classification by gene distribution: an overlapping subspace clustering approach

Area of Science:

• Genomics
• Bioinformatics
• Evolutionary Biology

Background:

• Horizontal gene transfer complicates evolutionary studies in lower organisms like bacteriophages.
• Traditional phylogenetic methods struggle with genomes exhibiting extensive gene exchange.
• Unsupervised clustering based on gene order and position offers a novel approach to genome classification.

Purpose of the Study:

• To develop an unsupervised overlapping subspace clustering algorithm for genome classification.
• To preserve genomic information by associating clusters with gene arrangement patterns.
• To identify multiple conserved patterns within genomes, particularly those affected by horizontal gene transfer.

Main Methods:

• A novel strategy to vectorize genomes based on gene distribution was employed.
• Existing subspace clustering and biclustering algorithms were evaluated.
• A generic subspace clustering algorithm (HARP) was extended to incorporate overlapping capabilities.

Main Results:

• The algorithm was applied to bacteriophage genomes, yielding groupings consistent with the Phage Proteomic Tree.
• Four conserved gene distribution patterns were identified across 441 phage genomes, centered on key genes like terminase and portal.
• A distinct subgroup of Sfi21-like phages with divergent genome organization was observed, leading to the identification of nine new phage members.

Conclusions:

• The proposed method objectively classifies genomes based on gene order, content, and position, aiding evolutionary studies.
• This approach is effective for genomes with high genetic exchange and conserved gene arrangements, as demonstrated in bacteriophages.