An artificial intelligence system developed by MIT University researchers in the US has found a powerful new antibiotic (“halicin”), which can kill many types of bacteria resistant to other antibiotics such as carbapenem.
It is the first time artificial intelligence has found a new antibiotic from scratch and even powerful, opening up new possibilities in the field of pharmacy, as something similar could be done in the future for other types of drugs, e.g. for neurodegenerative diseases or cancer.
Scientists hope that the “smart” system will be able to design several new drugs, based on what it has learned about chemical structures that allow drugs to kill bacteria.
Using a new machine learning algorithm, scientists and engineers, led by Professor of Medical Engineering & Science James Collins and Professor of Electrical & Computer Engineering Regina Barzilai, who published the paper in the biology journal Cell, analyzed several potential chemicals, eventually finding one with strong antibiotic features.
In recent decades few new antibiotics have been found and most are only slightly altered versions of pre-existing drugs. Current methods of developing new antibiotics are very expensive, time consuming and usually limited to a small range of chemicals.
“We are facing an increasing crisis due to an increasing number of pathogenic microorganisms that are becoming resistant to existing antibiotics and an anemic discovery of new antibiotics by pharmaceutical and biotechnological companies. If we do not deal with the crisis by 2050, deaths each year due to infections from resistant bacteria will reach ten million, more than annual cancer deaths.”
Collins said.
Artificial intelligence is the new supporter in this effort, as it promises to find completely new antibiotics, faster and at a much lower cost. The new antibiotic, tested in the laboratory, killed several strains of bacteria that are resistant to all known antibiotics (Clostridium difficile, Acinetobacter baumannii, Mycobacterium tuberculosis, Enterobacteriaceae, etc.), with the exception of the resistant aerobic Pseudomonas aeruginosa.
It also “cleaned up” within 24 hours infections in laboratory animals (mice) that were infected with acinetobacter baumannii, a highly resistant bacterium that has infected many American soldiers in Iraq and Afghanistan.
Furthermore, the well-known E.coli was found not to develop resistance to the new antibiotic after a 30-day treatment with it. On the contrary, this micro-organism develops resistance e.g. to the antibiotic ciprofloxacin of the quinolon class within only one to three days.
“We have developed a platform that allows us to rein in the power of artificial intelligence to pave the way for a new era in the discovery of new antibiotics. So we have already discovered an impressive molecule, which is arguably one of the most powerful antibiotics ever discovered.’
Collins mentioned.
The initial “training” of the algorithm was done through its feed with approximately 2,500 molecules of active substances, of which 1,700 were medicines approved by the competent U.S. supervisory authority, the Food and Drug Administration (FDA), while the remaining 800 were natural products.
After its “education”, the “smart” system was tested on about 6,000 other molecules with a possible antibiotic action contained in the Drug Repurposing Hub database. Of these -within just a few hours- artificial intelligence selected 100 candidates for physical tests and eventually a molecule with a chemical structure different from any known antibiotic was found to be the most effective against E.coli.
This substance, named “halicin”, referring to the famous artificial intelligence system HAL, of the book and the film “2001: Space Odyssey”, had in the past been considered as a possible anti-diabetic drug, but without success.
The researchers plan to do further tests of “halicin” in collaboration with a pharmaceutical company or nonprofit organization, so that they eventually develop an antibiotic for use in humans.
Also the same artificial intelligence system, searching a database (ZINC15) with 107 million molecules, brought to light another 23 potential active antimicrobials.
After animal testing, scientists have reached eight with antibacterial properties and two of them in particular are promising to form antibiotics, so they will be tested further.
See the scientific publication at the following address:
https://www.cell.com/cell/fulltext/S0092-8674(20)30102-1;utm_medium=homepage
Source:
(Συνολικές Επισκέψεις: / Total Visits: 1)
(Σημερινές Επισκέψεις: / Today's Visits: 1)