MIT of Boston, the software has selected the best molecules, among seven thousand candidates, to combat a bacterium feared in hospitals, Acinetobacter baumannii, resistant to many of the drugs available today
Artificial intelligence has enormous potential in the medical field. An important confirmation comes from a study by the Massachusetts Institute of Technology (Mit) of Cambridge, of theHarvard University and ofMcMaster University from Hamilton, Canada, published in the magazine Nature Chemical Biology
. What is surprising is not only the discovery itself (a molecule effective against a superbug), but also the procedure with which the result was achieved. Research into new antibiotics has been slow in recent decades, while the phenomenon of resistance, i.e. the ability of some bacteria to escape existing therapies, is clearly increasing. The possibility of using artificial neural networks (i.e. machine learning models) to select thousands of molecules allows us to reach goals that were unimaginable just a few years ago.
The authors of the new study examined, thanks to artificial intelligence, approx 7 thousand molecules evaluating their effectiveness againstAcinetobacter baumannii
, a Gram-negative bacterium that resists almost totally (about 90%) to some of the most powerful antibiotics available (fluoroquinolones, aminoglycosides, carbapenems). Furthermore capable of incorporating DNA fragments into its genome from the environment or other organisms, thus increasing its potency. It survives for a long time on surfaces and for this reason poses a huge danger in hospitals: if an immunocompromised patient is infected with the bacterium, he may develop an infection that his immune system is unable to fight and for which there are no effective therapies. American and Canadian researchers, thanks to a broad drug repurposing hub (database of molecules), have identified a possible new weapon: it is theabaucin, a compound with activity directed specifically against A. baumannii
Mechanism of action
By analyzing the thousands of molecules, artificial intelligence (deep learning) identified in a couple of hours those potentially active against the bacterium. The researchers then narrowed the selection criteria, arriving at a list of 240 chemical compounds, which was then further narrowed down to 9. Until the best candidate is discovered, that is RS102895 (renamed abaucin), originally created as a potential antidiabetic drug. Interesting study for various reasons — he comments Francis Scaglioneprofessor of Pharmacology at the University of Milan and clinical pharmacologist at the Niguarda hospital – especially for the mechanism of action of the molecule, which is new: abaucin would be capable of removing energy from the bacterium, blocking its activities. All antibiotics developed to date work differently.
Furthermore, according to the study authors, candidate RS102895 could only strike A. baumannii and not other types of bacteria – an advantageous feature in fighting the worrying phenomenon of antibiotic resistance, which arises precisely from wide-ranging therapies (moreover often used improperly). We don’t know if abaucin will become a drug – concludes Scaglione -, because most of the compounds identified by artificial intelligence fail to pass clinical trials on humans. It works against the superbug, but it will be needed now evaluate any side effects and pharmacokinetics (i.e. the effect it has on the body), including the level of absorption. However, the discovery is important, also because once a new mechanism of action has been identified modifications can be made to the selected molecule to improve its performance.
May 28, 2023 (change May 28, 2023 | 07:45)
© REPRODUCTION RESERVED
#Artificial #intelligence #discovered #antibiotic #hours