Table 1 Features used in this study for the prediction of pVOG-pVOG functional association

Co-occurrence

Calculated by dividing the number of genomes where both pVOGs have hits by the total number of genomes where either of the pVOGs have hits (Jaccard similarity)

Average distance

Minimum distance in nucleotides of the alignment envelopes (sensu hmmsearch) between the two pVOGs, averaged across all common genomes

Co-orientation

Fraction of hits where both pVOGs are found on the same strand

Convergent orientation

Fraction of hits where both pVOGs are found on opposite strands with 3’ ends facing each other

Divergent orientation

Fraction of hits where both pVOGs are found on opposite strands with 5’ ends facing each other

Mean ANI

Average Nucleotide Identity (ANI) of all genome pairs where the pVOGs co-occur, calculated with fastANI [26]

Mean AAI

Average Amino acid Identity (AAI) of all genome pairs where the pVOGs co-occur, calculated with CompareM [27]