Aim 4

Optimize antibiotic leads through metabolomic profiling of toxicity, efficacy, and mechanism of action

Preclinical antibiotic leads fail in clinical trials primarily due to low efficacy, high toxicity, or poor pharmacokinetics. Lead optimization is an important phase of drug discovery that attempts to optimize these parameters, primarily through testing in animal models. However, animal testing is slow, expensive, and worst of all, a poor predictor of clinical trial success.

To overcome these limitations in this project, we will pursue human “organ-on-chip metabolomics”, to screen human tissue samples directly for the efficacy and toxicity of antibiotic leads. This approach allows screening to be done accurately, cheaply, and at scale, providing an information-rich dataset of the metabolic responses of human tissues to different antibiotic candidates.

Using AI-based approaches to analyze the complex metabolomic data from these “organs-on-chip”, we also expect to develop a more comprehensive understanding of the mechanisms of action of different drugs by observing their broader effects on tissues. Feeding back these data into the active-learning drug design models will further train them on the intended and unintended metabolic effects of different drugs on human tissues, which may optimize further drug predictions with regard to toxicity and side effects.

The enhanced throughput of the organ-on-chip approach compared to animal models will allow for early triage of compounds that are toxic to humans, streamlining the overall drug development pipeline.