Protein structure-informed bacteriophage genome annotation with Phold.

Area of Science:

• * Bioinformatics
• * Structural Biology
• * Virology

Background:

• * Bacteriophage (phage) genome annotation is crucial for understanding viral function and therapeutic potential.
• * Current methods rely on sequence homology, which can limit accuracy and sensitivity.

Purpose of the Study:

• * Introduce Phold, a novel annotation framework leveraging protein structural information.
• * Improve the accuracy, speed, and scalability of phage genome functional annotation.

Main Methods:

• * Phold integrates the ProstT5 protein language model and Foldseek structural alignment tool.
• * Utilizes a database of over 1.36 million predicted phage protein structures with functional labels.
• * Benchmarked against existing sequence-based homology approaches.

Main Results:

• * Phold demonstrates superior functional annotation sensitivity compared to sequence-based methods.
• * Achieves consistent annotation of over 50% of genes in average phage genomes and 40% in archaeal virus genomes.
• * Phage proteins frequently exhibit structural homology to conserved proteins across the tree of life, especially those involved in nucleic acid metabolism and enzymatic functions.

Conclusions:

• * Phold offers a robust and efficient solution for phage genome annotation.
• * Its structure-based approach enhances functional prediction and broadens understanding of phage biology.
• * The tool supports the advancement of phage-based therapeutics and research.