Data and code for: Deduplication reshapes the perceived family composition of mobilized colistin resistance (mcr) genes - a curated census, phylogeny and classification audit

Abstract

Curated dataset and analysis code for a deduplicated census of the mobilized colistin resistance (mcr) gene family in the August 2026 NCBI snapshot (AMRFinderPlus Reference Gene Catalog 2026-08-07.1). Contents. Four archives: (1) core curated datasets (1,298 → 1,106 records, 454 unique proteins, 284-protein QC-triage core set; identity matrix, clusters, multiple-sequence alignments, maximum-likelihood and neighbour-joining trees, domain matrix, taxonomy tables); (2) all 21 figures in four formats (PNG 600 dpi, vector PDF, text-preserving SVG, native-editable PPTX); (3) the full analysis script set; (4) REPRODUCE.md with a per-file SHA-256 manifest (repro_manifest.json). Headline content. By unique protein, mcr-3 (202) rather than mcr-1 (84) is the largest family - a 2.4-fold reversal of the record-level ranking - reproduced under threshold-matched proxy implementations of AMRFinderPlus, CARD/RGI and ResFinder across all 75 cells of their combined identity × coverage grid. Nineteen divergent candidates retain the PF08019+PF00884 dual-domain architecture, share only 41.3-68.9% identity with their nearest named family, and are recalled by none of the three systems. Five families are monophyletic; the five catalytic residues are invariant in 284/284 sequences; a recombination signal is reported without coordinates for mcr-3 because the breakpoint positions failed replication across four disjoint subsets. Version anchors. AMRFinderPlus catalog 2026-08-07.1; NCBI snapshot August 2026; CARD and ResFinder at commit level. License. CC BY 4.0. Please cite the accompanying paper and this DOI.