Authors: Heather Jones, Future Science Group
By harnessing the power of artificial intelligence, researchers from the Massachusetts Institute of Technology (MIT; MA, USA) have identified a powerful new antibiotic compound. The study, published in Cell, could allow for the discovery of many potential new antibiotics to combat the rising tide of antibiotic resistance.
Recent decades have seen a decline in the number of new antibiotics appearing on the market, and those that have been approved show great similarity to existing drugs. Drug development is hindered by the high cost of screening methods, length of time and the narrow spectrum of chemical diversity.
“We’re facing a growing crisis around antibiotic resistance, and this situation is being generated by both an increasing number of pathogens becoming resistant to existing antibiotics, and an anemic pipeline in the biotech and pharmaceutical industries for new antibiotics,” commented James Collins (MIT).
In the search for completely novel compounds, a team from MIT incorporated a machine-learning model trained to analyze the molecular structures of compounds to uncover any antibacterial properties.
“The machine learning model can explore, in silico, large chemical spaces that can be prohibitively expensive for traditional experimental approaches,” explained Regina Barzilay (MIT).
Unlike previous models, these neural networks are able to learn molecule representations automatically and map them into continuous vectors that are subsequently used to predict their properties.
In this study, the researchers trained the model on approximately 2500 molecules, including 1700 US FDA-approved drugs, and 800 natural products exhibiting a wide range of structures and bioactivities.
After training the model, the team used it to screen approximately 6000 compounds. The model identified one molecule that had both strong antibacterial properties and a unique chemical structure. Further testing revealed it to have low toxicity to human cells.
The researchers subsequently named the molecule halicin in reference to ‘2001: A Space Odyssey’ and tested it against several isolated bacterial strains. Halicin was effective against almost every strain, many of which are resistant to other antibiotics, including Clostridium difficile, Acinetobacter baumannii and Mycobacterium tuberculosis.
Following the initial success, the team used halicin to treat mice infected with an antibiotic-resistant strain of A. baumannii. The novel drug cleared the infection within 24 hours.
Further research indicated that halicin works by disrupting the electrochemical gradient across bacteria cell membranes. Researchers believe that this mechanism of destruction could be difficult for bacteria to develop resistance to.
“When you’re dealing with a molecule that likely associates with membrane components, a cell can’t necessarily acquire a single mutation or a couple of mutations to change the chemistry of the outer membrane,” commented Jonathan Stokes (MIT).
Testing halicin on E. coli revealed a stark difference between halicin and the antibiotic ciproflaxin, with no resistance to halicin observed within a 30-day period compared with resistance developed within 1–3 days for ciproflaxin.
After further screening, the researchers identified eight other molecules that were structurally different and showed antibacterial activity.
The team plans to pursue further studies of halicin and use their model to design new antibiotics and to optimize existing molecules.
You might also like:
Source: Stokes JM, Yang K, Swanson K et al. A deep learning approach to antibiotic discovery. Cell 180(4), 688–702 (2020); http://news.mit.edu/2020/artificial-intelligence-identifies-new-antibiotic-0220