Data and code for: Pilus architecture, not mobile-element or defence gene content, marks host-jump propensity in the zoonotic pathogen Streptococcus agalactiae

Analysis code, curated reference markers with UniProt accessions, per-genome feature matrices, lineage labels, model reports, retrospective-panel scores, geNomad corroboration, and software parameters supporting the study "Pilus architecture, not mobile-element or defence gene content, marks host-jump propensity in the zoonotic pathogen Streptococcus agalactiae." Host-jump propensity was modelled as a latent lineage trait across 1,157 public group B Streptococcus genomes resolving into 56 lineages across fish, cattle and human hosts, using lineage-blocked cross-validation and a lineage-level label-permutation test. The deposited files reproduce every reported metric, including the primary pilus-and-mobilome model (lineage-blocked AUC-ROC 0.75, permutation P 0.003) and the retrospective panel. All analysed genome assemblies are public and their accessions are included.