Taxonomy-aware, sequence similarity ranking reliably predicts phage-host relationships

Abstract

BackgroundSimilar regions in virus and host genomes provide strong evidence for phage-host interaction, and BLAST is one of the leading tools to predict prokaryotic hosts from phage sequences. However, BLAST-based host prediction has three major limitations: (i) top-scoring sequences do not always point to the actual host, (ii) mosaic phage genomes may match to many, typically related, bacteria, and (iii) phage and host sequences may diverge beyond the point where their relationship can be detected by a BLAST alignment. ResultsWe created an extension to BLAST, named Phirbo, that improves host prediction quality beyond what is obtainable from standard BLAST searches. The tool harnesses information concerning sequence similarity and bacteria relatedness to predict phage-host interactions. Phirbo was evaluated on three benchmark sets of known phage-host pairs, and it improved precision and recall by 11-40 percentage points over currently available, state-of-the-art, alignment-based, alignment-free, and machine learning host prediction tools. Moreover, the discriminatory power of Phirbo for the recognition of phage-host relationships surpassed the results of other tools by at least 10 percentage points (Area Under the Curve = 0.95), yielding a mean host prediction accuracy of 57% and 68% at the genus and family levels respectively, and drops by 12 percentage points when using only a fraction of phage genome sequences (3 kb). Finally, we provide insights into a repertoire of protein and ncRNA genes that are shared between phages and hosts and may be prone to horizontal transfer during infection. ConclusionsOur results suggest that Phirbo is a simple and effective tool for predicting phage host relationships.