COLOMBO

COLOMBO predicts genomic islands (GIs) in microbial genomes to identify horizontally transferred regions and infer potential gene donors.

Key Features:

• Genomic island prediction: Identifies large DNA segments consistent with horizontal gene transfer (HGT) and flags putatively alien genes within microbial genomes.
• SIGI-HMM plugin: Implements SIGI-HMM to predict GIs and infer potential donors for each alien gene based on gene-level codon usage comparisons.
• Codon usage (CU) analysis: Compares each gene’s CU to a curated set of CU tables representing microbial donors or highly expressed genes.
• Inhomogeneous hidden Markov model (HMM): Determines states and emission probabilities at the gene level and uses transition probabilities derived from classical test theory with a sensitivity controller to refine predictions.
• Statistical testing and masking: Employs multiple statistical tests to identify putatively alien genes and to mask genes likely to be highly expressed.
• Plugin architecture: Modular design supports integration of analytical plugins such as SIGI-HMM.
• Implementation and formats: Implemented in Java, accepts EMBL-formatted input and produces GFF output for integration with genome analysis pipelines.
• Benchmarking and validation: Predictions have been benchmarked against annotated GIs and other prediction methods, showing consistency with known findings.

Scientific Applications:

• HGT characterization: Delineates horizontally transferred genomic regions to study mechanisms and impacts of horizontal gene transfer.
• Microbial evolution and adaptation studies: Facilitates analysis of genomic determinants underlying rapid microbial adaptation and niche acquisition.
• Donor inference: Supports inference of potential donor taxa for individual alien genes based on codon usage similarity.
• Genomic island annotation and comparison: Enables annotation of GIs and comparison of predictions to existing annotations and alternative methods.
• Gene origin hypothesis generation: Provides evidence to generate hypotheses about the origin and potential function of horizontally acquired genes.

Methodology:

SIGI-HMM performs gene-level codon usage (CU) comparisons to curated CU tables, applies multiple statistical tests to call putatively alien genes and mask highly expressed genes, estimates states and emission probabilities for an inhomogeneous HMM with transition probabilities derived from classical test theory and a sensitivity controller, and is implemented in Java with EMBL input and GFF output.