Harnessing machine learning to develop new antibiotics for Neisseria gonorrhoeae

PROJECT SUMMARY/ABSTRACT This proposal presents a five-year research and career development plan focused on machine learning (ML)-guided antibiotic development for multidrug-resistant Neisseria gonorrhoeae. The candidate, Melis Anahtar, MD, PhD, has completed clinical pathology training at Massachusetts General Hospital (MGH) and is

now a postdoctoral researcher under the mentorship of Dr. James Collins at MIT and the Broad Institute. Dr. Collins pioneered the field of ML-based antibiotic discovery and has trained over 70 independent research investigators. The proposed project builds on Dr. Anahtar’s previous research experiences in the vaginal

microbiome and in genomic antimicrobial resistance (AMR) determinants. AMR poses an existential threat by rendering existing antibiotics useless. This threat is exemplified by N. gonorrhoeae, a sexually transmitted gram-negative bacterium that afflicts ~87 million people every year and has recently developed resistance to

ceftriaxone, the last remaining highly effective antibiotic treatment. Untreated gonorrhea causes infertility, pregnancy complications, neonatal blindness, severe disseminated infection, and death, prompting the CDC and WHO to recognize drug-resistant N. gonorrhoeae as one of the five most urgent AMR threats to human

health. Traditional drug discovery has failed to keep up with AMR and there is a critical need for innovative approaches to fill the antibiotic development pipeline with promising candidates. To solve this problem, Dr. Anahtar proposes using predictions from ML models to efficiently guide focused testing and development of

novel compounds for the treatment of drug-resistant N. gonorrhoeae. The central hypothesis is that ML models such as graph neural networks can identify new antibiotic candidates for N. gonorrhoeae by predicting growth inhibitory activity from chemical structures given high-quality training data. In preliminary work, Dr. Anahtar

phenotypically screened a foundational library of 38,000 chemically diverse compounds for N. gonorrhoeae growth inhibition, used the results to train an ML model, and deployed the model to identify novel compounds with activity against multidrug-resistant N. gonorrhoeae. To expand the pool of novel antibiotic candidates, Dr.

Anahtar will use this ML model to screen large (10^8) chemical libraries in silico and then validate the predictions in vitro (Aim 1). The mechanism of action of the most potent and non-toxic compounds will then be determined using microscopic, proteomic, and genomic approaches (Aim 2). Finally, their in vivo efficacy will

be tested in a preclinical animal model of N. gonorrhoeae infection (Aim 3). The long-term goal is to discover new biological insights into the treatment of drug-resistant bacteria. This work will be performed at the Broad Institute, MIT, and MGH, which provide an exceptional training environment and ample scientific resources.

The candidate is supported by an outstanding scientific advisory committee with decades of experience in N. gonorrhoeae, AMR, medicinal chemistry, and mentoring physician-scientists. This career development award will enable Dr. Anahtar’s transition to independence with a focus on understanding and combating AMR.