Data and code for: Antimicrobial resistance determinants in starter cultures and probiotics: a mobility-aware genomic survey of 8,407 assemblies across seven species

This deposit accompanies a manuscript on whether antimicrobial resistance determinants carried by starter cultures and probiotics are actually mobile, that is, whether they sit next to a transposable element in the genome that carries them, and on how far that answer depends on the thresholds used to call mobility and on the quality of the assemblies screened. It contains: (1) the accession list and the NCBI assembly metadata behind every analysis, including the assembly level, size and GC content of all 8,407 assemblies (01_metadata/); (2) the per-determinant mobility assignments, the mobile-element BLASTX hits behind them, the 27-combination threshold-sensitivity grid, and the two negative controls, namely random genomic positions and a matched 3 kb window (03_mobility/); (3) the per-gene, per-species and per-strain tables behind every reported number, including the source data of Figures 1 to 7 and of Supplementary Tables S1 to S15 (05_tables/); (4) the average-nucleotide-identity clustering of the seven species, in both the representative-anchored and single-linkage variants, with the per-assembly N50 table (07_strains/); (5) the curated mobile-element protein set used as the BLASTX reference (envs/db/); (6) the analysis scripts that regenerate the tables and figures, together with a standard-library verifier for the manifest (scripts/); and (7) documentation: a README with the pipeline, the order to run it and a reproduction report, a table-to-source map, a script-to-outputs map, and glossaries of every column and every coded cell value. The rendered artwork is not shipped; every figure is redrawn from its shipped source table by scripts/make_figures.py. Version 2 corrects an accession-parsing error in scripts/build_anchored_clusters.sh: filenames whose assembly name is not `_ASM...` (for example GCA_000442825.1_LaDSM20242V1) were previously dropped, so 25 L. acidophilus genomes had no ANI edges and the anchored clustering over-split. The corrected clustering yields 2,164 anchored clusters (version 1 reported 3,081) and 1,708 single-linkage clusters; Table S14 has been recomputed accordingly (anchored representative 87/257 = 33.9%, any-member 115/309 = 37.2%; single-linkage 74/232 and 91/264). The single-linkage representative flag was also corrected to one genome per cluster (highest N50). No other data, tables or scripts changed. The 8,407 genome assemblies are public in NCBI GenBank/RefSeq under the accessions listed in 01_metadata/P09_accessions.txt and are not redistributed here. The per-genome AMRFinderPlus annotation is not redistributed either: it is regenerated by scripts/scan840.sh with AMRFinderPlus v4.2.7 and Reference Gene Catalog 2026-08-07.1, both recorded in metadata.json. MANIFEST.csv lists every file with its size and SHA-256. Code is released under the MIT License and the data, tables and documentation under CC BY 4.0, as stated in the LICENSE file.