2024-01-11 12:05:26
The European Center for Disease Control and Prevention (ECDC) estimates that 33,000 deaths a year in Europe result from antibiotic-resistant bacteria. While few new classes of antibiotics have been discovered or synthesized over the past 60 years, a team of 21 researchers, led by Felix Wong and James J. Collins, announce the discovery of a new class of antibiotics using AI and explainable deep learning to tackle the highly pathogenic Staphylococcus aureus.
Felix Wong is a postdoctoral fellow in the laboratory of James J. Collins, the Termeer Professor of Medical Engineering and Science at MIT. The former is a co-founder of Integrated Biosciences while the latter is the founding chairman of the company’s scientific advisory board. The article reporting their work entitled “Discovery of a structural class of antibiotics with explainable deep learning” was published in Nature on December 20.
Co-authors include researchers from the Massachusetts Institute of Technology (MIT), the Broad Institute of MIT and Harvard, the Wyss Institute for Biologically Inspired Engineering, and the Leibniz Institute for Polymer Research in Dresden, Germany.
Antibiotic resistance of Staphylococcus aureus
In this study, the researchers focused on the discovery of structural classes of antibiotics effective against Staphylococcus aureus, the Latin name for Staphylococcus aureus, a potentially fatal bacterium, particularly feared in hospitals, certain strains of which are particularly resistant to antibiotics.
They first screened an original set of 39,312 compounds containing most known antibiotics, natural products and structurally diverse molecules. The team assessed their antibiotic activities and cytotoxicity profiles, then used ensembles of graphical neural networks to predict antibiotic activity and cytotoxicity for an additional 12,076,365 compounds.
The researchers used explanatory graphical algorithms, designed to analyze the molecular structure of compounds in a way that makes the prediction process more understandable and explicit. They thus identified the substructures responsible for the predicted antibiotic activity, having a low probability of toxicity for human cells.
They then tested 283 compounds empirically and were able to identify a structural class exhibiting high selectivity, capable of overcoming resistance, and possessing favorable toxicological and chemical characteristics.
To evaluate the effectiveness of the new class of antibiotics against Staphylococcus aureus (MRSA) under conditions similar to those of a human infection, the researchers carried out in vivo experiments on mice. They found a significant reduction in the number of bacteria, demonstrating the effectiveness of this structural class in the topical and systemic treatment of MRSA in mice.
These results suggest that this class of compounds could be promising for the development of new antibiotics. However, their effects in humans must be verified; clinical studies are already underway.
Felix Wong says:
“This discovery of a new class of antibiotics is a groundbreaking result that shows that artificial intelligence and explainable deep learning are the only ones capable of catalyzing drug discovery. Our work makes several very powerful models available to the public to accurately predict the activity and toxicity of antibiotics. Importantly, this is one of the first demonstrations that deep learning models can explain what they predict, with immediate and profound implications for how drug discovery is carried out and about how effectively we can find new drugs using AI”.
Article references:
“Discovery of a structural class of antibiotics with explainable deep learning” Nature, https://doi.org/10.1038/s41586-023-06887-8
Auteurs :
Felix Wong, Erica J. Zheng, Jacqueline A. Valeri, Nina M. Donghia, Melis N. Anahtar, Satotaka Omori, Alicia Li, Andrew Cubillos-Ruiz, Aarti Krishnan, Wengong Jin, Abigail L. Manson, Jens Friedrichs, Ralf Helbig , Behnoush Hajian, David K. Fiejtek, Florence F. Wagner, Holly H. Soutter, Ashlee M. Earl, Jonathan M. Stokes, Lars D. Renner, James J. Collins.
1704979231
#Researchers #discover #class #antibiotics #deep #learning