Antimicrobial Resistance Co-occurrence Networks Exhibit Robust-Yet-Fragile Topology: A Network-Theoretic Basis for Hub-Targeted Combination Therapy

Antibiotic cycling — the sequential rotation of antibiotic classes — is structurally equivalent to random node removal in a scale-free network. AMR gene co-occurrence networks exhibit robust-yet-fragile (RYF) topology: random removal (cycling) preserves network connectivity while targeted hub removal causes rapid collapse. This provides a network-theoretic rationale for hub-targeted combination therapy over cycling strategies. We construct a 42-node, 73-edge AMR co-occurrence network calibrated to WHO GLASS 2022 and CARD co-occurrence priors and show that at 48% node removal, targeted hub removal reduces the largest connected component to 14% versus 67% under random removal (4.8-fold difference). blaKPC-3 (betweenness centrality 0.42) is identified as the highest-value structural target. Four exploitable collateral sensitivity pairs are curated from published clinical data. Code and data: https://github.com/AegisMindApp/amr-ryf-networks