AlphaFold 3 proteomic modelling reveals multiple photosystem structural homologs in freshwater cyanophages

Abstract

Accurate protein structure prediction followed by structural homology detection enable the functional annotation of large numbers of otherwise obscure viral protein-coding genes. Here we employ AlphaFold 3 modeling and DALI structural homology search to predict the structures and functions of all 219 cryptic open reading frames in two representative freshwater cyanophages. We discover 28 previously unknown structural homologs, including three putative PsaD proteins in the same cyanophage, several viral structural proteins, and (to our knowledge) the first reported virus-encoded cyanobacteriochrome (DALI Z-score > 4.2 in all cases). Our results suggest that photosystem proteins may be more widespread in freshwater cyanophages than previously thought and emphasize the importance of applying structural homology detection methods when annotating viral genomes.