Novel Antibiotic Discovered Using the Power of AI
7th June 2023 - Last modified 19th October 2023
By Pete Cussell PhD, Science Writer
Researchers have harnessed deep learning technology to discover a powerful new antibiotic that could prove critical in the fight against a deadly superbug.
A new study published last week in Nature Chemical Biology [1] used an AI-guided approach to identify a novel antibiotic that proves highly effective against deadly hospital superbug Acinetobacter baumannii. The bug in question has been classified as a ‘critical’ threat by the World Health Organisation, who placed A.baumannii on its list of ‘priority pathogens’ – essentially a ‘most wanted’ list that includes those bacteria that pose the greatest risk to human health, and for which new antibiotics are sorely needed [2].
The promising study, carried out by researchers from McMaster University and the Massachusetts Institute of Technology, has sparked a great deal of interest – see here, here and here for just some examples of the stories that hit the news last week.
Amid a backdrop of ever-growing global incidences of multidrug-resistant superbugs and a scarcity of effective antimicrobial agents to treat them, it is hoped that the AI-led approach pioneered in the study could finally address the critical need for effective antibiotics following decades of stagnation in the market.
The Growing Threat of Antibiotic Resistance
Antibiotic resistance refers to the ability of certain bacteria to survive and proliferate in the presence of antibiotic agents that would typically kill them or inhibit their growth. It’s a substantial and growing issue because it renders the antibiotics available to us much less effective, making resistant infections highly challenging – and sometimes impossible to treat.
Bacteria become resistant as a result of genetic changes, either through mutation following exposure to an antibiotic agent, or by acquiring pro-resistance genes from parent bacteria. These genetic alterations allow the bacteria to develop mechanisms and phenotypic traits to overcome an antibiotic drug’s mechanism of action. Resistance has been exacerbated over the years by overuse, inappropriate prescribing and extensive agricultural application of broad-spectrum antibiotic agents.
Antibiotic-resistant bacterial strains are no new threat. The first resistant strain was described in 1940, when a penicillin-resistant Staphylococcus was identified. Soon after, antibiotics would have a ‘boom’ from the late 1960’s through to the early 1980’s, an era which saw a multitude of new antibiotics developed to overcome the resistance problem. After this time though, the pipeline soon dried up, with fewer and fewer novel antibiotics being developed, giving resistant bacterial strains the opportunity to flourish [3].
A.baumannii is one of the most threatening resistant strains at present – it’s most prominent in critically ill patients on mechanical ventilation or with central venous catheters, and most commonly causes pneumonia or bloodstream infections. Worryingly, many strains of Acinetobacter have developed resistance to all antibiotics, including carbapenems, which are known as the antibiotic of last resort. The bacteria can survive for prolonged periods of time on surfaces including shared equipment, enabling the strain to spread rapidly in a healthcare setting [4].
AI to the Rescue?
Faced with the challenge of identifying novel antibiotics to overcome the increasing problem of multidrug resistant bacteria, a team of US researchers hit the headlines following the development of an AI neural network with the ability to bioinformatically screen thousands of antibiotic molecules, and discover new structural antibiotic classes. The model operates by exchanging local chemistry data between adjacent atoms and bonds to determine the likelihood of a molecule having antimicrobial properties.
Once the deep learning model had been trained using known growth-inhibiting compounds, the team set to work analysing some 6,680 candidate molecules. Following the screen, 240 of the promising lead molecules were tested for antibiotic properties in the lab. Of the 240 tested, 9 of the molecules passed the researcher’s threshold of >80% growth inhibition. Next, the team discounted any molecules with structural features similar to known antibiotic classes and those with potentially dangerous nonspecific activity. This led the team to identify one exciting lead molecule in particular, named abaucin.
The team tested abaucin against 41 different strains of A.baumannii and observed that the compound could overcome all of its intrinsic and acquired resistance mechanisms. The researchers finally tested abaucin on a wound infection mouse model, and showed that the molecule was able to supress A.baumannii infection. These findings are remarkable, and this highlights the vast potential of machine learning for the discovery of fundamentally novel antibiotic drugs.
A Bright Future Ahead
It’s clear that novel antibiotics are sorely needed, and this exciting news demonstrates the significant value that AI could bring to antibiotic discovery. While broad-spectrum antibiotics of the past are easily overcome by bacteria, AI methods enable researchers to identify antibiotics specific to target bacterial strains. Furthermore, this can be done at a massively increased rate and at a fraction of a cost compared with traditional drug screening methods. The discovery of aubacin could be seen as a proof of concept to researchers and pharmaceutical companies, and with hope, could trigger a new ‘boom’ in antibiotic development.
At Alto, we love science news! Talking about it, reading about it – and of course writing about it. To find out more about our science writing expertise, contact us!
References
(1) Liu, G., Catacutan, D.B., Rathod, K. et al. Deep learning-guided discovery of an antibiotic targeting Acinetobacter baumannii. Nat Chem Biol (2023).
(3) Ventola C. L. (2015). The antibiotic resistance crisis: part 1: causes and threats. P & T, 40(4), 277–283.
(4) Vázquez-López, R., Solano-Gálvez, S. G., Juárez Vignon-Whaley et al. (2020). Acinetobacter baumannii Resistance: A Real Challenge for Clinicians. Antibiotics, 9(4), 205.
