FARM dataset v1.0

Companion dataset for the FARM model (Drug-conditioned cross-attention for antimicrobial-resistance prediction). Code repository: https://github.com/Ruheng-W/FARM Web server: https://cdc.biohpc.swmed.edu/farm/intro License: GNU General Public License v3.0 (GPL-3.0). See LICENSE and COPYRIGHT in this record. Release date: 2026 This record bundles the harmonised genotype–phenotype training data, the independent out-of-distribution screen, all reference files (drug SMILES, KO-cluster annotations, drug-class taxonomy), and the splits used to produce every reported result. Note on coverage. The manuscript reports results on 10 cohorts totalling 7,216 isolates. 6 of those cohorts are redistributed here under GPL-3.0; 4 clinical cohorts (CF, Chile, Rabin, TIDB) are subject to data-use restrictions imposed by the contributing institutions and are not included in this record. Their per-cohort dimensions are documented in ten_cohorts/COHORT_MANIFEST.md, and the underlying data are available from the corresponding authors on reasonable request. See Section 5 for details.