We are living in a post-antibiotic era. The miracle drugs that shaped our modern healthcare system are losing effectiveness, and we aren't developing enough new drugs to meet patient needs. Innovative strategies to kill resistant bacteria and protect our antibiotic arsenal are urgently needed. Biodegradable polyanhydride nanoparticle-based nanomedicines could shape the next generation of antimicrobial therapies by improving drug biodistribution and targeting. By controlling drug release and improving antibiotic potency, they provide dose sparing and dose frequency reduction capabilities that could improve compliance. Unfortunately, traditional screening approaches are impractical for nanomedicines due to the massive physicochemical dataspace contributed by drug, polymer, and nanoparticle properties. Additionally, the relationships between these properties and nanomedicine efficacy are complex and nonlinear, impeding first principles modeling. Improved methods of screening and modeling antimicrobial nanomedicines are needed to realize their full potential.

In this dissertation, we report the adaptation of a high-throughput method to rapidly screen a novel polyanhydride copolymer chemical space for interesting drug delivery properties. These CPTEG:SA copolymers were shown to erode more rapidly than traditional polyanhydride copolymers, resulting in rapid, chemistry-dependent drug release within three days. This rapid release could be beneficial for fast-growing pathogens by providing a quick, suppressive antibiotic dose that is maintained over several days. The high-throughput method was adapted to synthesize polymer-drug films to screen for thermodynamic mixing interactions that influence release kinetics.

These methods enabled high-throughput synthesis of antibiotic-loaded nanoparticle libraries, which were evaluated for their controlled release capabilities and antimicrobial efficacy against the opportunistic, resistant pathogen Burkholderia cepacia. Multivariate data analytics approaches were used to identify key polymer, drug, and nanoparticle properties that determine nanomedicine release kinetics and efficacy. Graph theory was used to reduce the dimensionality of the descriptor data while preserving nonlinear relationships between formulations, enabling interrogation of nanomedicine design pathways. The dimensionally compressed descriptor space was used to develop predictive models for nanomedicine release kinetics and efficacy. These models successfully predicted the release kinetics and encapsulation efficiency for nanoparticles encapsulating two drugs not included in the training data set. In terms of efficacy, the model successfully predicted whether untested individual nanomedicines or nanomedicine cocktails would provide improved potency over soluble drug.

From these results, we proposed two informatics-assisted frameworks to rapidly screen nanomedicine candidates capable of killing resistant bacteria and controlling drug release. Overall, this dissertation has provided the first steps toward a broader framework for the rational design of antimicrobial nanomedicines. Lead candidates identified by this framework could provide new therapies against resistant pathogens and enable repurposing of existing antibiotics limited by resistance.