Predicting antimicrobial mechanism-of-action from transcriptomes: A generalizable explainable artificial intelligence approach

Abstract

To better combat the expansion of antibiotic resistance in pathogens, new compounds, particularly those with novel mechanisms-of-action [MOA], represent a major research priority in biomedical science. However, rediscovery of known antibiotics demonstrates a need for approaches that accurately identify potential novelty with higher throughput and reduced labor. Here we describe an explainable artificial intelligence classification methodology that emphasizes prediction performance and human interpretability by using a Hierarchical Ensemble of Classifiers model optimized with a novel feature selection algorithm called Clairvoyance; collectively referred to as a CoHEC model. We evaluated our methods using whole transcriptome responses from Escherichia coli challenged with 41 FDA-approved antibiotics and 9 crude extracts while depositing 306 transcriptomes. Our CoHEC model can properly predict the primary MOA of previously unobserved compounds in both purified forms and crude extracts at an accuracy above 99%, while also correctly identifying darobactin, a newly discovered antibiotic, as having a novel MOA. In addition, we deploy our methods on a recent E. coli transcriptomics dataset in a different strain and a Mycobacterium smegmatis metabolomics timeseries dataset and showcase exceptionally high performance; improving upon the performance metrics of the original publications. We not only provide insight into the biological interpretation of our model but also that the concept of MOA is a non-discrete heuristic with diverse effects for different compounds within the same MOA, suggesting substantial antibiotic diversity awaiting discovery within existing MOA.

Interactions between fecal gut microbiome, enteric pathogens, and energy regulating hormones among acutely malnourished rural Gambian children

“…Used in the context aggregate networks and predictive modeling Protein energy malnutrition PEM Protein-energy malnutrition defined as a range of pathological conditions arising from inadequate calories and/or protein intake. [1] Modeling Clairvoyance Feature selection algorithm leveraged for phenotype-discriminative community detection [58] Hierarchical Ensemble of Classifiers HEC Graphical model where each internal node is a customized sub-model classifier with a unique feature set [58] Sub-model Machine-learning classification model used as internal node in a HEC model Leave subject out cross-validation LSOCV Cross-validation designed to simulate performance on a new subject This study Networks Background network BN Networks created from individuals who were WN for all This study Perturbed background network PBN Networks created when adding in a query individual to the background network This study Sample-specific network SSN Network with unique properties for each sample [80] Sample-specific perturbation network SSPN Network created from perturbation between BN and SSPN distributions This study Aggregate network AN Networks created from fitted sub-model coefficients This study Node The discrete objects within a network Edge Weighted connections between nodes Edge weight Association or perturbation strength of edge Perturbation Change in association strength of an edge between SSN and BN distributions Connectivity k ...…”

“…Clairvoyance feature selection and LSOCV accuracy have been adapted from Espinoza and Dupont et al. 2021 which were developed to model antimicrobial mechanism-of-action [58] .…”

A Silent Operon of Photorhabdus luminescens Encodes a Prodrug Mimic of GTP

“…The prediction based on this model revealed high uncertainty along the decision path with exceptionally high variance and low probability at the terminal prediction. This model has been used previously to flag the novelty of darobactin ( 29 ), which exhibits a variance profile similar to that of ADG, further suggesting either a single new target or multiple targets ( Fig. 5 ).…”

Elucidating the Mechanisms of Action of Antimicrobial Agents

“…Using DNA microarrays and RNA sequencing, gene expression studies have enabled glimpses into the response of bacteria to antibiotic ( 47 – 49 ). Like with morphology in bacterial cytological profiling, transcriptome signatures can be generated across a wide range of antibiotics to experimentally classify an unknown antimicrobial’s MOA.…”