Fold first, ask later: structure-informed function annotation of Pseudomonas phage proteins

Area of Science:

• Virology
• Structural Biology
• Bioinformatics

Background:

• Bacteriophages (phages) possess diverse proteins with potential applications.
• Functional elucidation of these phage proteins is challenging, limiting their exploitation.
• Hypothetical proteins from Pseudomonas-infecting phages represent a largely uncharacterized resource.

Purpose of the Study:

• To investigate structure-informed strategies for annotating hypothetical phage proteins.
• To assess the efficacy of combining different annotation methods.
• To create a valuable resource of predicted structures and annotations for phage proteins.

Main Methods:

• Curated a dataset of over 10,000 phage proteins from NCBI.
• Predicted protein structures using ColabFold.
• Assessed structural similarity using FoldSeek against PDB, AlphaFold, and Phold databases.
• Evaluated sequence-based (Pharokka) and structure-based annotation methods.

Main Results:

• Up to 43% of unannotated proteins were functionally annotated by integrating structure-informed approaches with UniProt data.
• Demonstrated the complementarity of various databases (PDB, AlphaFold, Phold) for annotation.
• Highlighted the importance of filtering for annotation quality.

Conclusions:

• Structure-informed methods are crucial for annotating hypothetical phage proteins.
• Combining diverse databases and rigorous filtering optimizes annotation pipelines.
• This study provides a resource and insights for exploring phage biology and applications.