The spread of antibiotic resistance through bacterial populations all over the world adds serious risk to many clinical procedures. This project builds on recent work applying graph neural network models to identify antimicrobial activity in known molecules, as well as new datasets from large combinatorial screens of antibiotics combined with other potentiating molecules. I will find ways to represent combinations of molecules as input to train models to predict intermolecular synergies, possibly incorporating genetic information about bacterial strains as well. Time permitting, we hope to validate our results empirically in new experiments by testing candidate pairs of molecules against major bacterial pathogens.

I’ve really enjoyed exploring antibiotic resistance, both in this lab and a gap year working in industry. The problem is urgent and complex, and many of the topics I’ve learned in biology and computation classes converge on it. I’m looking forward to getting more hands-on experience with machine learning and exposure to the biochemistry side of the field, and especially excited to work with the fantastic people in this lab again.