AI Discovers New Antibiotic to Combat Drug-Resistant Infections


Researchers from MIT and McMaster University have harnessed the power of artificial intelligence (AI) to uncover a groundbreaking drug that shows potential in fighting drug-resistant infections. Through the implementation of a machine-learning algorithm, the team successfully identified an antibiotic capable of eradicating Acinetobacter baumannii, a bacterial strain responsible for numerous drug-resistant infections, including pneumonia, meningitis, and other severe ailments. This breakthrough could mark a significant turning point in the battle against antibiotic resistance.

Acinetobacter baumannii is a pervasive species of bacteria frequently found in hospital environments, posing a threat to both patients and healthcare workers. Additionally, it is a leading cause of infections among injured soldiers in conflict zones such as Iraq and Afghanistan. In recent years, the rise of antibiotic-resistant bacteria has outpaced the development of new antibiotics, intensifying the need for innovative solutions to combat these resilient pathogens.

MIT announced that their researchers employed a machine-learning model to analyze a catalog comprising approximately 7,000 potential drug compounds. The model was trained to evaluate each chemical compound’s efficacy in inhibiting the growth of Acinetobacter baumannii. To obtain training data for the model, the scientists exposed the bacteria to roughly 7,500 distinct chemical compounds in a laboratory setting, observing their effects on inhibiting microbial growth. The model was then fed the molecular structures of each compound, enabling it to distinguish between structures that could effectively suppress bacterial growth.

After training, the model was deployed to analyze a separate set of 6,680 compounds that it had not encountered previously. Through rigorous experimentation, the researchers narrowed down the candidates to 240 hits, focusing on compounds with unique structures, distinct from existing antibiotics and the training data molecules. Following further testing, nine antibiotics emerged, including one that exhibited exceptional potency.

Remarkably, the antibiotic compound that stood out was initially explored for its potential as a diabetes drug. Researchers named the compound abaucin, and subsequent studies conducted on mice showcased its efficacy in treating wound infections caused by Acinetobacter baumannii. Lab tests also demonstrated its effectiveness against various drug-resistant strains of the bacteria isolated from human patients. Furthermore, additional experiments revealed that abaucin works by disrupting the process of lipoprotein trafficking, which is vital for protein transportation from the cell’s interior to its envelope.

The narrow spectrum of abaucin’s killing ability is regarded as a desirable trait, as it reduces the risk of bacteria rapidly developing resistance to the drug. Moreover, it is expected to spare the beneficial bacteria residing in the human gut, which play a crucial role in suppressing opportunistic infections.

Currently, a laboratory at McMaster University is dedicated to optimizing the medicinal properties of abaucin, with the ultimate goal of developing it for clinical use. Additionally, the authors of the study intend to employ their modeling approach to identify potential antibiotics for combating other types of drug-resistant infections.

The remarkable findings of this research were published on Thursday in the journal “Nature Chemical Biology,” shedding light on the immense potential of artificial intelligence in revolutionizing the search for new antibiotics to combat drug-resistant bacteria.

(YWN World Headquarters – NYC)


  1. The term AI is being used here for publicity purposes as this is no more than a regular type of computer program that ha been used for decades to extrapolate pieces of data from large databases. Something that is standard. The only difference here was the amount of data that was fed to the program.
    No program is any more “intelligent” than the person designing and writing the program. What makes the difference is the speed with which the computer can sift through large amounts of data. So called AI is merely a computer program that will do in a minute that which would take an ordinary human years to do and the ability to make as many mistakes in a minute that would take years for an ordinary human o accomplish. When using ChatGPT you’ll find the error frequency in the answers mind blowing.

  2. @silentmoshe, you are partially right about the hype. But in this case, this might be a true Machine Learning (ML) application rather than just fast computations. ML looks at a small number of cases (7,000 sounds small to me in this case) and generalizes them to other unseen cases. That is similar to a kid who sees a couple of dogs and now can recognize other dogs of the breeds he never saw before. This ability of (some) of the ML algorithms to generalize is fascinating and is still behind the human abilities in the same area.

    This is on contrast with less imaginative “AI” hype cases that require millions of training examples and then predict something similar to what it was trained on.