Determining the ototoxic potential of anti-microbial and anti-inflammatory therapeutics using machine learning and in vivo approaches

Creighton University, Omaha NE

Investigators

Linked publications & trials

Abstract

There are over 2,000 clinical trials for influenza and other microbial diseases, representing hundreds of drugs and drug combinations. While this pace of drug development is necessary for public health, it also comes with increased risk of producing therapies with significant side-effects. One likely side-effect of some drugs is permanent hearing loss. The potential of drugs to cause hearing loss is typically unassessed during drug development or clinical trials. Our research will rapidly assess anti-microbial and anti-inflammatory drugs for ototoxic potential. Microbial infections are often accompanied by inflammation, hence the selection of antiinflammatory drugs in additional to anti-microbial compounds. Our goal is to promote the development of safe therapeutics with minimal side effects. Some drugs used for controlling infections are associated with hearing loss, yet the ototoxic potential of these drugs is often based on individual case reports or in vitro experiments so the true ototoxic burden is largely unknown. We should not use patients as a testbed for a life-altering negative side-effect when there are rapid low-cost alternatives available. The objective of this proposal is to rapidly identify the ototoxic potential of anti-microbial and anti-inflammatory therapeutics using both in silico and in vivo approaches. We will achieve this objective with three Specific Aims: 1) Predict ototoxic potential of anti-microbial and anti-inflammatory therapies using Machine Learning (ML), 2) Determine the hair cell toxicity of anti-microbial and anti-inflammatory drugs in the zebrafish lateral line, and 3) Determine the degree to which predicted ototoxins cause hearing loss in rats. Our innovative ML model correctly categorizes ototoxins vs. non-ototoxins with 87% accuracy. In this project we will employ our current model for immediate ototoxicity detection and further optimize the model for better predictive accuracy. In parallel with the ML model, we will screen therapeutics in the larval zebrafish lateral line, which is an excellent platform to rapidly detect candidate ototoxins. Finally, we will validate predicted ototoxins from Aims 1 and 2 in rats using both physiological and morphological assays. Our research is significant because we will determine the ototoxic potential of new or repurposed antimicrobials and will inform efforts to 1) advance effective candidates that are not ototoxic, 2) modify successful yet ototoxic drugs and/or develop otoprotective co-therapies to preserve therapeutic efficacy while minimizing ototoxic side-effects, and 3) determine which patients require audiometric monitoring due to the drugs they receive. Our team includes experts in ototoxicity, medicinal chemistry, biostatistics, ML, and clinical expertise in large-scale human trials. This innovative approach will develop, train, and test a novel ML model to predict drug-related toxicity.