Combinatorial therapeutic approaches are an imperative to improve cancer treatment, because it is critical to impede compensatory signaling mechanisms that can engender drug resistance to individual targeted drugs. Currently approved drug combinations result largely from empirical clinical experience and cover only a small fraction of a vast therapeutic space. Here we present a computational network biology approach, based on pathway cross-talk inhibition, to discover new synergistic drug combinations for breast cancer treatment. In silico analysis identified 390 novel anticancer drug pairs belonging to 10 drug classes that are likely to diminish pathway cross-talk and display synergistic antitumor effects. Ten novel drug combinations were validated experimentally, and seven of these exhibited synergy in human breast cancer cell lines. In particular, we found that one novel combination, pairing the estrogen response modifier raloxifene with the c-Met/VEGFR2 kinase inhibitor cabozantinib, dramatically potentiated the drugs' individual antitumor effects in a mouse model of breast cancer. When compared with high-throughput combinatorial studies without computational prioritization, our approach offers a significant advance capable of uncovering broad-spectrum utility across many cancer types.