Researchers at the Massachusetts Institute of Technology have used brute computer power in tangent with artificial intelligence to identify a potentially revolutionary new antibiotic. Using machine learning, a derived subset of AI, scientists are able to sort through millions of pharmaceutical options for medications, helping achieve financial and chemical success way faster than conventional standards.
The idea behind implementing machine learning in the process of antibiotic chemicalization has been under public imagination for a great while now — that a computer algorithm would have the ability to triangulate which medications work best against which pathogens. Recently, scientists have finally developed neural networks that can contextualize what molecules do and don’t work against.
According to a statement by MIT on the matter, “In this case, the researchers designed their model to look for chemical features that make molecules effective at killing E. coli….To do so, they trained the model on about 2,500 molecules, including about 1,700 FDA-approved drugs and a set of 800 natural products with diverse structures and a wide range of bioactivities.”
This essentially implies that the algorithm was able to identify the varying chemical features of each E. coli effective killer, including some that are unconventional in relation to the patterned thinking of the human brain. The scientists compiled their algorithm in a database of 6,000 identified and named chemical compounds.
According to the researchers, the algorithm shook one particular molecule loose: a molecule referred to by the name halicin. The researchers claim that it was previously applied in an unsuccessful trial to treat diabetes, but in clinical tests, it showed immediate success among a variety of pathogens and bacteria. Additionally, it is relatively nontoxic to human cells as well.
Halicin disrupts the cellular equilibrium that allows cells to be physically held together. It dissolves bacteria from the outside by disabling their ability to preserve this electromagnetic equilibrium. This differs from how traditional medicine works, as these types of medicines physically destroy the cells. Researchers claim that this small mechanism variation is most likely why the new molecule can kill bacteria that other antibiotics cannot.
Within the lab, the research subjected halicin to dozens of bacterial strains, including a few that are known to be resistant to antibiotics: Clostridium difficile, Acinetobacter baumannii, and Mycobacterium tuberculosis. Halicin worked on all three of these in petri-dish testing, and killed all three species with the exception of one difficult lung pathogen.
The team used the AI-discovered model to uncover 23 additional molecules that could have antibiotic abilities as well.
What do you think? Does the stretch of AI into healthcare propose a newly astounding field of development tailored to the creation of new and improved pharmaceutical techniques? Or is it a stretch not worth taking due to the intricacy and large field of error possible in the algorithmic process? Thoughts?