Phirbo

Phirbo predicts prokaryotic hosts for phage (meta)genomic sequences by combining sequence similarity with bacterial relatedness through taxonomy-aware ranked comparisons using the Rank-Biased Overlap (RBO) measure to mitigate limitations of BLAST-based methods.

Key Features:

• Taxonomy-Aware Sequence Similarity Ranking: Incorporates taxonomy-aware rankings of sequence similarity and compares ranked lists of varying lengths using the Rank-Biased Overlap (RBO) measure, assigning greater importance to higher-ranking matches.
• Mitigation of BLAST Limitations: Addresses cases where top-scoring BLAST hits do not indicate the actual host, mosaic phage genomes match multiple bacteria, and divergence prevents detectable phage-host alignments.
• Improved Precision and Recall: Integrates sequence similarity with bacterial relatedness to improve precision and recall by 11–40 percentage points compared to alignment-based, alignment-free, and machine learning methods.
• Discriminatory Power and Accuracy: Demonstrates an Area Under the Curve (AUC) of 0.95 with mean host prediction accuracy of 57% at the genus level and 68% at the family level, retaining robust performance with only partial phage sequences (3 kb) with a 12 percentage point accuracy drop.
• Insights into Shared Genes: Identifies protein and non-coding RNA (ncRNA) genes shared between phages and hosts to provide information relevant to horizontal gene transfer during infection.

Scientific Applications:

• Viral ecology and host specificity: Predicts phage-host relationships to inform studies of viral ecology and host range in microbial communities.
• Microbiology and genomics: Supports investigation of bacterial relatedness and phage interactions in microbiology and comparative genomics analyses.
• Horizontal gene transfer studies: Enables examination of protein and ncRNA sharing between phages and hosts to study mechanisms of horizontal gene transfer and evolutionary dynamics.

Methodology:

Compares taxonomy-aware ranked lists of sequence similarity between phage and bacterial genomes using the Rank-Biased Overlap (RBO) measure, weighting higher-ranking matches more strongly.