PhageTailFinder

PhageTailFinder predicts phage tail-related proteins and identifies tail protein clusters to inform studies of phage host range and infection mechanisms.

Key Features:

• Algorithmic Foundation: Uses a two-state hidden Markov model (HMM) that leverages the modularity of phage tail proteins and does not rely solely on amino acid properties or secondary structures.
• High Predictive Accuracy: Achieved a true-positive prediction rate exceeding 80% for 571 phages in an evaluation involving 13 well-characterized phages and 992 complete phages from the NCBI database.
• ROC Values: Reports a general model ROC of 0.877, a morphologic model ROC of 0.968, and a median ROC exceeding 0.75 for novel phages across 992 complete phages.
• Application to Metagenomic Data: Maintained a ROC value of 0.895 when applied to 189,680 viral genomes from 11,810 bulk metagenomic human stool samples.
• Cluster Identification: Identifies tail protein clusters using the density-based spatial clustering of applications with noise (DBSCAN) algorithm.

Scientific Applications:

• Phage therapy research: Predicting and annotating phage tail proteins supports development and selection of phages as alternatives to antibiotics.
• Host-range and infection mechanism studies: Annotation of tail modules aids interpretation of phage host range and infection mechanisms.

Methodology:

Computational methods explicitly include a two-state hidden Markov model (HMM) for tail-probability prediction, sequence-independent modeling that accounts for the modular nature of tail proteins, and DBSCAN for tail protein cluster identification.