Gene drives are genetic constructs with super-Mendelian inheritance that can spread engineered alleles in wild populations to suppress disease vectors and invasive species, with deployment programs to begin within a decade. We developed a network-based model of gene drive spread with an arbitrary number of populations connected by gene flow. Using generative network models, we identified network centrality measures of release sites that can predict the deployment outcome. We show that population structure can determine gene drive spread dynamics, and identify aspects of population structure that should be measured in target populations of gene drive deployment programs. Such models are crucial to understand how gene drives are expected to spread in order to design safe deployment programs.