Antibiotic resistance is one of the world’s most severe public health threats. Many infectious disease-causing bacteria are resistant or becoming resistant to our current arsenal of antibiotics, and it is becoming increasingly difficult and expensive to discover new antibiotics. Traditionally, antibiotics were discovered from “natural” sources like soil bacteria or fungi, but in recent years, the results are becoming repetitive, with known antibiotics being repeatedly discovered. Modern methods in antibiotics discovery involve costly large-scale screening of millions of molecules, but they have not yielded any effective antibiotics for the past three decades. In order to combat this worrying problem, a group of scientists from MIT are using artificial intelligence to discover new and effective antibiotics.
Artificial intelligence has actually been used in drug design for decades, but it has not produced accurate results and thus has not been particularly useful. However, there have been many recent advancements in machine learning algorithms, such as the development of deep neural networks: an algorithm that is roughly designed to work like the human brain. Scientists first trained the model with 2,335 molecules that are either anti-bacterial or not. Then, the algorithm looks at each molecule and classifies them automatically. By applying the trained model to another 107 million molecules, it was able to predict a powerful new antibiotic compound, halicin.
They tested halicin with a panel of antibiotic-resistant bacteria, and found that halicin is able to successfully kill many dangerous pathogens including Mycobacterium tuberculosis, the causative agent of tuberculosis. Additionally, halicin kills these bacteria in a novel way that is very different from other known antibiotics, thus emphasizing its potential to kill bacteria that are resistant to known antibiotics. Even after 30 days of continuous treatment, E. coli did not develop any resistance to halicin. Therefore, halicin has proven to be extremely effective in killing bacteria and avoiding resistance in bacteria.
At the end of the study, the researchers were able to use this same algorithm to predict a couple more antibiotic compounds that they want to test further in the future. The entire process of predicting potential antibiotics and then testing them took a couple of weeks, while a traditional antibiotic discovery process might take years. This underscores the remarkable speed at which new antibiotic compounds could be predicted and tested using artificial intelligence. Given the current slow and expensive process of antibiotic development, artificial intelligence could be the key in winning our race against the increasing spread of antibiotic resistance.
Managing Correspondent: Wei Li
Original Article: A Deep Learning Approach to Antibiotic Discovery. Cell.
News Article: Powerful antibiotic discovered using machine learning for first time. The Guardian.
Powerful “Space Odyssey” Antibiotic Developed by AI. Technology Networks.