North American researchers have identified a molecule – abaucine - that appears to be able to fight one of the most drug-resistant bacteria in hospital environments, Acinetobacter baumannii by training an AI programme.
Antibiotic resistance is one of the great health emergencies of the present and near future, as the World Health Organisation has repeatedly stated, echoing numerous health and scientific authorities that for years have been recording a worrying increase in resistant species, in the light of increasingly less effective pharmacological methods.
To fight this, significant help could come from artificial intelligence (AI), which can carry out research in a short time that, until now, took months or years, and was never as thorough or complete. This was shown in a study published in the scientific journal Nature Chemical Biology, in which a team of pharmacologists and bioengineers from McMaster University in Hamilton, Canada, and the Massachusetts Institute of Technology in Boston, searched for effective drugs against one of the most resistant bacterial strains to traditional antibiotics (and therefore most responsible for deadly infections): Acinetobacter baumannii.
One of the favourite habitats of this micro-organism is hospitals, where it can survive classic disinfection systems for weeks (by acquiring increasingly resistant functions), not least because it has a special ability to acquire and transfer fragments of genetic material from other species of bacteria, which it incorporates into its DNA. Among those most often acquired are the very genetic code traits that produce antibiotic resistance. This bacterium is also able to create a kind of protective structure on its outer surface, which makes it even more difficult to treat successfully. The result is pneumonia, meningitis and wound infections that are often impossible to treat in the most vulnerable patients.
An AI programme ‘trained’ directly by researchers
To combat this specific bacterium, North American researchers have developed a special AI programme with all possible information about the structure and vulnerability of Acinobacter baumannii so that the computer system can 'learn' the most effective strategy to combat it. The researchers then compared 7,500 possible drug structures and tested what the algorithm was able to predict.
The final result was a single molecule, called abaucine, which did not disappoint expectations, neither in in vitro cultures nor in animals with wounds infected with A. baumannii. The infectiologists confirmed that aubacine acts by disturbing the metabolism of certain fat-related proteins, typical of the bacterium. If new tests confirm these results, trials could begin on the first human volunteers.
Millions of molecules to be studied
Two more general aspects, which signal a change in approach, should be emphasised. In contrast to what has generally been done so far, in the case of Acinetobacter baumannii an antibiotic with as specific a spectrum of action as possible was sought, to be used only in the event of an infection caused by this type of bacterium. While this course action may be a limitation (because in severe cases several different infective species often coexist) it also reduces the risk of resistance (promoted, on the other hand, by more 'generic' antibiotics).
It should also be noted that, according to the researchers, one could conceivably find new molecules from archives of millions of structures, not only using an algorithm, as with abaucine, but also randomly, thanks to the power of AI.
This means performing tasks that would usually take years in a few hours. No human being could ever work on even remotely comparable amounts of information in such a short time. Therefore, in this case, the advantages are truly significant, and the risks non-existent.
