The problem of the rapid spread of antibiotic resistance among pathogenic bacteria is one of the most pressing problems of modern medicine, so the development of new antibiotics is now a very important task. Recently, in the journal Cell, American researchers reported that they were able to find a new potential broad-spectrum antibiotic in the Drug Repurposing Hub’s compound database using machine learning. The discovered substance was named galicin. The authors of the work experimentally showed that galicin has bactericidal activity against bacteria of different phylogenetic groups, including such human pathogens as the causative agent of tuberculosis Mycobacterium tuberculosis and the causative agent of colitis Clostridioides difficile. Our article focuses on a new strategy for finding potential antibiotics using machine learning.

In the second half of the 20th century, in connection with the discovery of penicillin, streptomycin and other antibiotics, it began to seem that infectious diseases no longer pose a serious threat to humanity. The situation began to change when it became clear that pathogenic bacteria surprisingly quickly acquire the mechanisms that provide them with resistance to antibiotics from the arsenal of doctors. Moreover, due to horizontal gene transfer, which is widespread among prokaryotes, they exchange genes that provide protection against antibiotics at an amazing speed.

The situation is further complicated by the fact that the search for antibiotics has practically reached a dead end in recent years. Most of the antibiotics currently used in clinical practice were detected by screening for secondary metabolites released by soil microorganisms. In this way, β-lactam antibiotics, aminoglycosides, polymyxins and glycopeptides were discovered. A number of antibiotics are of semi-synthetic origin: they were obtained by hanging various chemical groups on already known antibiotics. There are also fully synthetic antibiotics in the arsenal of doctors, for example, antibiotics of the pyrimidine and quinolone series.

However, progress in the development of new antibiotics has become frustratingly low in recent decades. Screening for natural metabolites most often leads to re-detection of already known antibiotics. Attempts to obtain new antibiotics, to which bacteria have not developed resistance, by means of chemical modification of known antibiotics, are also not successful. According to some projections, if the current situation with the development of new antibiotics does not change, by 2050 infectious diseases will account for 10 million deaths annually [1]. Is there a way out of this crisis?