An important step in achieving pharmaceutical independence from China - repurpose existing drugs with AI for new uses
http://www.digitaljournal.com/tech-and-science/technology/op-ed-a-i-halicin-molecular-antibiotics-major-mit-success/article/567760
Op-Ed: 'AI ‘Halicin’ molecular antibiotics major MIT success
By Paul Wallis Feb 25, 2020 in Technology
Boston - Research by MIT has applied AI to antibiotics, and it’s speaking volumes about this line of attack. A new AI-based project has found low-toxicity molecular antibiotics that kill some of the world’s worst bacteria.
Tests indicate that the new antibiotic, (called “Halicin” after Hal the computer in 2001 A Space Odyssey) killed even MRSA bacteria resistant to all known antibiotics. Even more impressively, Halicin was originally developed as a diabetes treatment, so it wasn’t really on the radar as an antibiotic.
The MIT AI predicted the effect of the antibiotic, which disrupts mitochondrial function in pathogens, effectively depriving them of energy. (Mitochondria are cell “batteries”, which power cellular functions.)
Halicin even cleared up a particularly nasty infection called A. baumanni, which has been infecting troops in Afghanistan. A topical application of ointment containing Halicin eliminated the infection in 24 hours. (An ointment???!!! Wow. That’s an achievement. From major problem to “smear some of this on you” is no minor effort.) Some diseases of this type are truly pernicious, able to last for years after initial infection.
The anti-mitochondrial approach may well be a good broad-spectrum approach to treating infections. Low toxicity adds excellent value when treating at-risk patients who may have compromised immune systems or other serious conditions reducing their ability to recover. These conditions make infections so much more dangerous. A general “kill the bugs anywhere anytime” capability could be an excellent all-round asset for patient management and recovery.
AI proves its point
This is an indicator of the inherent values of AI and machine learning in a whole new context. This high-value research can now be done in days, not years. AI can process and assess hundreds of millions of compounds quickly.
Medical research is the classic Big Data environment. It’s tough. There’s so much data generated. That vast weight of information, with a certain irony, makes it hard to access and use the data. It’s very difficult to get the full benefits of it by the old methods, simply due to the need to organise and process vast ranges of disparate data.
Antibiotics have been effectively stalled for years and fighting antibiotic-resistant diseases has been a pretty thankless task. The sheer scale and range of the problem, with multiple potentially serious diseases emerging, has been working against researchers. AI research allows for much more intensive analysis of possible compounds.
More to the point – The logic is different, too. Machine learning clearly isn’t just learning by rote. This is a very significant departure from simply developing similar drugs. Halicin is a very different type of drug, unlike other antibiotics. The AI obviously wasn’t looking at classes or types of drug, but functional options. The mitochondrial line of attack is deadly and could have many more ramifications in treating whole classes of diseases which have that mitochondrial vulnerability.
One way or another – This research has made so many points. It’s found an incredibly useful antibiotic, opened up a whole new method of managing some very dangerous conditions and truly horrible pathologies.
If Halicin is all it seems to be, AI has pretty much just paid for itself with this work alone. Let’s see what else it can do. This really is the future of artificial intelligence, and it’s looking good.