Combination Antibiotic-Focused Machine Learning Models for Integration into Experimental Workflows

Abstract

Antibiotic resistance is a growing crisis with far-reaching impacts on public and global health [1][2]. With antibiotic resistance impeding the efficacy of antibiotics and the discovery of new antibiotics slowing [2], combination therapies provide ways of extending the lifespan of existing antibiotics and treating resistant infections. Given the immense time and resources required to experimentally test combination therapies, computational strategies to accelerate the discovery of combination therapies are needed. In this project, machine learning models were developed to predict the synergistic activity of antibiotic-adjuvant pairwise combinations against E. coli. A variety of architectures were developed and tested. The most effective model type, a 2D Feed Forward Neural Network, was deployed through a Jupyter Notebook user interface for easy usage by experimentalists. Thus, this project engineers an accurate, computationally efficient, and user- friendly implementation of a machine learning algorithm that can integrate into experimental workflows to accelerate the discovery, screening, and validation of combination antibiotic therapies.