Bacteria are amazing creatures. They adapt rapidly to any stress that is put on them or at least some part of the population. Because of the careless use of antibiotics to treat illnesses where they are not effective, antibiotic resistance has arisen.
Some countries even allow the sale of antibiotics over the counter. The problem is getting out of control to the point that drug-resistant strains of bacteria that can no longer be treated are becoming a serious threat. About
35,000 people a year die from these infections.
New antibiotics are desperately needed.
Stokes et al. have developed a novel way of searching for antibiotics using machine learning. Using artificial intelligence programming they trained an algorithm to identify the types of molecules that kill bacteria by feeding it 2,500 drugs and natural compounds and comparing that to how well the compound killed
E. coli. The program learned the traits of a successful antibiotic. After training, they set the algorithm to screen 6,000 compounds to see if any of them had the traits that would make it a bacteria killer, and that is was unlike currently known antibiotics. After a few hours, the algorithm suggested several compounds. When tested, one of them proved to be very effective against known drug-resistant strains. The authors dubbed the compound Halicin in honor of the computer Hal in 2010 A Space Odessy.
The Guardian has an excellent summary of the work.