SSG-LUGIA

SSG-LUGIA identifies genomic islands (GIs) and horizontally transferred genes in prokaryotic genomes to localize regions of horizontal gene transfer and study bacterial evolutionary adaptation.

Key Features:

• Unsupervised anomaly detection: Employs an unsupervised learning approach based on anomaly detection techniques to predict GIs without requiring prior functional annotations.
• Compositional bias analysis: Leverages atypical compositional biases of alien genes to localize genomic islands.
• Signal processing integration: Integrates signals from signal processing literature to enhance detection of anomalous genomic regions.
• Draft genome compatibility: Applicable to draft and newly sequenced prokaryotic genomes for GI analysis.
• Benchmark evaluation: Evaluated on the IslandPick benchmark dataset and on 15 well-studied bacterial genomes.
• Case-study validation: Includes detailed analysis on Salmonella typhi CT18 and assessments on Corynebacterium diphtheriae NCTC13129 and Pseudomonas aeruginosa LESB58.
• Precision–recall performance: Demonstrated a more favorable balance between precision and recall compared to frequently used existing methods.

Scientific Applications:

• Genomic island localization: Localization of clusters of horizontally transferred genes (genomic islands) in prokaryotic genomes.
• Horizontal gene transfer detection: Identification of alien genes transferred via horizontal gene transfer.
• Evolutionary and adaptation studies: Investigation of bacterial evolutionary dynamics and adaptation across ecological niches.
• Analysis of draft and novel genomes: Characterization of genomic islands in newly sequenced or draft bacterial genomes.
• Method benchmarking: Comparative evaluation of GI prediction methods using benchmark datasets such as IslandPick.

Methodology:

Applies an unsupervised learning approach based on anomaly detection that leverages atypical compositional biases of alien genes and integrates signals from signal processing literature.