-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
Powerful new antibiotics identified thanks to AI
- Thread starter Deleted member 46948
- Start date
Time for general over production and use from farming to humans to livestock.
Lets go!
Like with any other promising technical advancement it will be blocked as long as possible by the corresponding lobbies that could lose money from this.
OP
OP
I didn't realize there's money in antibiotics not working. A brisk trade in dead bodies, huh?
So this is the pivot needed for the antibiotic startup phase 2. Most of the ones started in the 2000s are slowly going bust if not there already. One of the reasons why big pharma doesn't do antibiotics anymore is because it isn't a recurring business model. When the patient is cured, they won't need to purchase them. It's the reason why governments had been stepping up to find new antibiotics but if this is fast-tracked it might fill that hole that's coming up and soon (when some of these companies go bankrupt, which they are currently, the trademarked antibiotics developed by them will be in a tbd state. This would strain the supply to hospitals, clinics, and other medical organisations). So this couldn't come at a more opportune time.
I remember watching a show on why there's no antibiotic research because you only take them for a week or two. Pharmaceutical companies prefer to develop drugs you take for a long time, maybe for life.
It is important to understand the difference between there not being market incentive for developing new antibiotics (which is the case) and deciding not to do it for some nefarious purpose (not the case).
You know the answer to this and it's not good. Still it's nice to live in hope.
That's not really the problem here, the problem is that if a new antibiotic capable of tackling multi-resistant bacteria are found it's basically not going to be used. It's going to be kept in reserve and only used in tiny quantities for the worst cases in order to avoid bacteria getting resistant to it as well.
There's no money in that. Hence the need for public funding.
OP
OP
Yup, this.
Honestly I figure we only need to buy a couple of decades worth of time until we're able to solve a lot of problems that currently loom on the horizon via magic technology, but it's 50/50 whether we manage to destroy ourselves through some cataclysmic fuckup before that happens.
OP
OP
It sometimes feels like both of these reasonings are one and the same.
You just need a different pipeline for developing non-profitable pharmaceuticals. You can't force companies to develop things against their will, you can either incentivize it or fully fund it from public resources.
i don't really think there are a ton of new anti biotics so there is money to be made if you develop one for strains that are hard to treat. a lot of the current common anti biotics have generics so those aren't money makers.
Only if we massively scaled it up to an absurd level. But there's no current commercial application to justify it at the moment, nor in investing to increase the efficiency through scientific development.
Regardless, it'd take a very long time.
That's what I'm saying. Hopefully AI can have a breakthrough at extremely efficient ways to do it.
As someone who has been suffering from atypical tuberculosis since last september (heavy to treat, very resitent bacteria) all I can say is fuck those bacterias and fuck antibiotics that do shit to your body you do not want. It's a loose-loose situation.
This is good news though all around, maybe not for me but for some in a few years that are untreatable.
This is good news though all around, maybe not for me but for some in a few years that are untreatable.
OP
OP
It's basically the premise of Reassortment in Person of Interest.
A man of culture and taste, I see :)
It's an interesting story, but it's hard to understand what they're actually doing. The AI identifies molecules "by their activity", without focusing on "molecular groups". That seems logically impossible, since you can't know the activity until you've tested it. And you can't extrapolate activity unless you know the molecular groups. I suspect it just reflects the difficulty of explaining this in layman's terms and they've probably done something very clever (molecular groups probably means not caring about some specific high-level "drug family" assignments, rather than low level "beta unsaturated carbonyl" groups).
I expect a neural net AI could do a better job at finding correlations than a human, especially when you need to combined multiple properties (functional molecular groups, the right molecular weight and polarity to bypass various membranes, and other things relating to metabolism).
I'd like to see a paper on exactly what they did.
I expect a neural net AI could do a better job at finding correlations than a human, especially when you need to combined multiple properties (functional molecular groups, the right molecular weight and polarity to bypass various membranes, and other things relating to metabolism).
I'd like to see a paper on exactly what they did.
Pretty sure the paper is this one: https://www.sciencedirect.com/science/article/pii/S0092867420301021.
There's also a github page of their model (https://github.com/chemprop/chemprop), which describes the format of the data they used to train it.
