This dataset comprises genomic sequences of predicted prophage elements from Bacillus subtilis genomes, organized into separate FASTA files based on clustering results. A total of 64 FASTA files are provided. Within each file, individual prophage sequences are listed, each preceded by a header in the following format: >AP011541.2_Phage_7_Pos_3240205-3266407. This format denotes the host genome ID, phage identifier, and the prophage coordinates within the host genome. Each FASTA file is uniquely numbered to correspond with a viral Operational Taxonomic Unit (vOTU), representing the prophage group to which the sequences belong. The prophage elements were sourced from both complete B. subtilis genome sequences available in NCBI's database (191 genomes; last accessed March 2023) and from 40 local soil microscale isolates (PS strains; PRJNA437002). Prophage coordinates within the bacterial genomes were identified using the PHASTER web tool. Functional categories assigned by PHASTER (complete, questionable, incomplete) were disregarded, and all predicted prophage elements were included. These prophage elements were then clustered using the vContact2 tool. Briefly, prophage genes were predicted using Prodigal, translated into proteins, and subjected to reference-free phylogenetic clustering via vContact2. This tool constructs a network based on the number of shared proteins between phages. The ClusterOne algorithm was then applied to this network to infer groups of related phages, which were used to define vOTUs for further analysis. In addition to the vOTU-based clusters, two artificial clusters were created: Outliers and Singletons. Outliers share some proteins with different vOTUs but fall below the 20% similarity threshold, while Singletons represent prophage elements with no similarity to other clusters. Dataset is described in publication: doi: https://doi.org/10.1101/2024.07.03.601884