Cocrystal Network-Based Recommendation: A Multi-Objective Random Walk-Based Network Model for Virtual Screening of Drug–Drug Cocrystals

Drug-drug cocrystals (DDCs) are composed solely of active pharmaceutical ingredients, offering traditional cocrystal benefits and potential intercomponent synergistic therapeutic effects. However, systematic, efficient computational methods for DDC screening are lacking. To address this challenge, here we introduced CCNBR, a multiobjective random walk network model that combines cocrystal network topology with molecular structural features. By leveraging a third-order path-based weighted random walk algorithm, CCNBR effectively captures key supramolecular interactions that drive cocrystal formation, enabling rapid identification of promising DDC candidates. To validate its practical utility, we applied CCNBR to evaluate the cocrystal-forming potential among 15 commonly used antihypertensive drugs and experimentally tested 105 drug combinations. Ultimately, two drug pairs successfully formed cocrystals, ranking first and second on the CCNBR recommendation list. Notably, we successfully identified Furosemide-Telmisartan cocrystal, which not only exhibits complementary therapeutic mechanisms but also demonstrates significantly improved solubility and bioavailability in both in vitro and in vivo evaluations.