In the early twentieth century, the discovery of penicillin happened by chance, thanks to the intuition of a brilliant English microbiologist, Alexander Flemming, who, in 1928,for realized that the bacterial culture he was monitoring had been annihilated by the presence of a layer of mold.
Since then, medical research, conducted through the traditional empirical method, has allowed for the creation of numerous types of antibiotics. Despite this, also due the abuse of antibiotics, in recent years several studies have shown the proliferation of bacterial forms that are increasingly resistant to drugs and therefore have become more difficult to fight.
The Interagency Coordination Group (IACG), commissioned by the United Nations to study the resistance of bacteria to antibiotics, published in 2019 a report, in which, it was estimated that drug-resistant diseases could cause 10 million deaths per year by 2050.
Precisely to cope with the need to develop new antibiotics capable of fighting the most resistant strains of bacteria, researchers of the MIT (Massachusetts Institute of Technology) have conducted specific research, using Machine Learning techniques, that have led to the discovery of a new drug: halicin.
The name of the compound was chosen in honour of the Hal 9000 supercomputer from the film “2001: A Space Odyssey”.
Prof. James Collins, professor of Medical Engineering at MIT, explained that the activity was carried out in a clean slate mode, i.e. without providing the algorithm with starting various hypotheses.
The research was carried out by designing a model that had chemical characteristics, such that to make the molecules effective in killing the bacterium Escherichia coli. To do this, the model was trained on about 2,500 unique molecules, including about 1,700 FDA-approved drugs and a set of 800 natural products with different structures and a wide range of bioactivities.
Once the training phase was completed, the researchers tested the model on the Broad Institute’s Repurposing Hub, a library of about 6,000 curative compounds that are still undergoing clinical evaluation. Through in vitro testing of the compound, a molecule (renamed halicin) was identified that had strong antibacterial activity and a chemical structure different from any existing antibiotic. Researchers have also shown through further studies that this molecule is likely to have low toxicity to human cells. The peculiarity of halicin is also represented by its ability to act even against the so-called resistant strains, that is, it has managed to inhibit the attempts of mutation of bacteria resistant to normal antibiotics. Its bactericidal action has, in fact, been tested with excellent results against other bacteria resistant to traditional treatments, such as Clostridium difficile, Acinetobacter baumannii, and Mycobacterium tuberculosis.
Prof. Collins said that the purpose of the activity carried out was to “develop a platform that allows us to harness the power of artificial intelligence to usher in a new era of antibiotic drug discovery”.
Through Machine Learning, says Regina Barzilay, professor of electronic and computer engineering in the computer science and artificial intelligence laboratory at MIT (CSAIL) who collaborated on the research, “it is possible to explore large chemical spaces that can be prohibitive for traditional experimental approaches” .