The cause of and solution to all of life’s problems, circa 2025.
I think medical advancements are going to go crazy in the next few years. Let’s just keep an eye on AI just in case though.
Now this sounds like a good use of AI
Sweet. Strange to see actual good news.
The title is clickbait if the content is decent.
There have been something like 20 new antibiotics in the last 10 years that didn’t come from AI.
There are tons of new antibiotics, I’m not sure what this 60 years figure is coming from.
What the AI does is design theoretical molecules, not all of which are practical or even possible, a few of which may have some practical application. The value is the sheer speed at which the AI can generate these designs; humans still must do the actual analysis.
For all the poo-pooers who don’t read articles
They used AI to discover a new CLASS of antibiotics. They searched tested tens of thousands of compounds and did computational predictions from MILLIONS of compounds. This narrowed down to several hundred compounds for which they did further testing. They then found one class of compounds that have strong antibiotic properties and that have a different chemical structure compared to other existing antibiotics.
This is actually an incredible breakthrough and doesn’t deserve the shade that clueless commenters are dropping here.
>The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis1,2,3,4,5,6,7,8,9. Deep learning approaches have aided in exploring chemical spaces1,10,11,12,13,14,15; these typically use black box models and do not provide chemical insights. Here we reasoned that the chemical substructures associated with antibiotic activity learned by neural network models can be identified and used to predict structural classes of antibiotics. We tested this hypothesis by developing an explainable, substructure-based approach for the efficient, deep learning-guided exploration of chemical spaces. We determined the antibiotic activities and human cell cytotoxicity profiles of **39,312 compounds** and applied ensembles of graph neural networks to **predict antibiotic activity and cytotoxicity for 12,076,365 compounds**. Using explainable graph algorithms, we identified substructure-based rationales for compounds with high predicted antibiotic activity and low predicted cytotoxicity. We empirically tested 283 compounds and found that compounds exhibiting antibiotic activity against Staphylococcus aureus were enriched in putative structural classes arising from rationales. Of these structural classes of compounds, one is selective against methicillin-resistant S. aureus (MRSA) and vancomycin-resistant enterococci, evades substantial resistance, and reduces bacterial titres in mouse models of MRSA skin and systemic thigh infection. Our approach enables the deep learning-guided discovery of structural classes of antibiotics and demonstrates that machine learning models in drug discovery can be explainable, providing insights into the chemical substructures that underlie selective antibiotic activity.
First new class of antibiotics, so title is a bit misleading. There are several new antibiotics coming to market each year, albeit they are variations of existing structural families. Kind of exciting if AI can generate an entirely new class of structures. The real test is how well they perform in trials.
it’s secretly something that’ll wipe out humans because AI was like “gotta save the earth”
When I was doing my PhD, what often happens is that I accidentally found a good material and later made up a story saying I did machine learning first and that leads to the discovery, just to ride the AI hype and make my paper easier to publish. I really wished I didn’t have to do that because it is the discovery itself, not the AI, that really matters.
And that’s how the zombie plague began
Huh, some good news for a change. Refreshing.
We need to start using this new class of antibiotics en masse for pig farms ASAP. Those bacteria aren’t going to grow resistance themselves.
I actually can’t imagine *not* using machine learning extensively for anything like this these days. If you’re processing a lot of data, there is likely ML.
We don’t have *any* AI today though, it literally does not exist and everyone would for sure know about it if we did, so AI has not been used and the headline is factually wrong.
Big deal, get ready for the revolution
Indian doctors (heavy breathing): can’t wait to prescribe this for regular fever
Now let’s overprescribe it to people who don’t have bacterial infections /s
Can the AI give us healthcare?
How can they tell the antibiotics are using AI.
Time to push AI to finding effective treatments for more ailments. Hoping for effective treatments, vaccines, and more discoveries.
Does this shit work for cdiff?
This type of thing, and material science, are the 2 main things ML tech is going to be the greatest at. Grinding away at data to narrow down viable, interesting properties from nearly unlimited possibilities. Sorting through immense amounts of data for things is exactly what this technology is perfect for.
The next step of course is even more amazing, but current tech is still good enough to be perfect for this kind of thing.
23 comments
The cause of and solution to all of life’s problems, circa 2025.
I think medical advancements are going to go crazy in the next few years. Let’s just keep an eye on AI just in case though.
Now this sounds like a good use of AI
Sweet. Strange to see actual good news.
The title is clickbait if the content is decent.
There have been something like 20 new antibiotics in the last 10 years that didn’t come from AI.
There are tons of new antibiotics, I’m not sure what this 60 years figure is coming from.
What the AI does is design theoretical molecules, not all of which are practical or even possible, a few of which may have some practical application. The value is the sheer speed at which the AI can generate these designs; humans still must do the actual analysis.
For all the poo-pooers who don’t read articles
They used AI to discover a new CLASS of antibiotics. They searched tested tens of thousands of compounds and did computational predictions from MILLIONS of compounds. This narrowed down to several hundred compounds for which they did further testing. They then found one class of compounds that have strong antibiotic properties and that have a different chemical structure compared to other existing antibiotics.
This is actually an incredible breakthrough and doesn’t deserve the shade that clueless commenters are dropping here.
Original paper here https://www.nature.com/articles/s41586-023-06887-8
​
>The discovery of novel structural classes of antibiotics is urgently needed to address the ongoing antibiotic resistance crisis1,2,3,4,5,6,7,8,9. Deep learning approaches have aided in exploring chemical spaces1,10,11,12,13,14,15; these typically use black box models and do not provide chemical insights. Here we reasoned that the chemical substructures associated with antibiotic activity learned by neural network models can be identified and used to predict structural classes of antibiotics. We tested this hypothesis by developing an explainable, substructure-based approach for the efficient, deep learning-guided exploration of chemical spaces. We determined the antibiotic activities and human cell cytotoxicity profiles of **39,312 compounds** and applied ensembles of graph neural networks to **predict antibiotic activity and cytotoxicity for 12,076,365 compounds**. Using explainable graph algorithms, we identified substructure-based rationales for compounds with high predicted antibiotic activity and low predicted cytotoxicity. We empirically tested 283 compounds and found that compounds exhibiting antibiotic activity against Staphylococcus aureus were enriched in putative structural classes arising from rationales. Of these structural classes of compounds, one is selective against methicillin-resistant S. aureus (MRSA) and vancomycin-resistant enterococci, evades substantial resistance, and reduces bacterial titres in mouse models of MRSA skin and systemic thigh infection. Our approach enables the deep learning-guided discovery of structural classes of antibiotics and demonstrates that machine learning models in drug discovery can be explainable, providing insights into the chemical substructures that underlie selective antibiotic activity.
First new class of antibiotics, so title is a bit misleading. There are several new antibiotics coming to market each year, albeit they are variations of existing structural families. Kind of exciting if AI can generate an entirely new class of structures. The real test is how well they perform in trials.
it’s secretly something that’ll wipe out humans because AI was like “gotta save the earth”
When I was doing my PhD, what often happens is that I accidentally found a good material and later made up a story saying I did machine learning first and that leads to the discovery, just to ride the AI hype and make my paper easier to publish. I really wished I didn’t have to do that because it is the discovery itself, not the AI, that really matters.
And that’s how the zombie plague began
Huh, some good news for a change. Refreshing.
We need to start using this new class of antibiotics en masse for pig farms ASAP. Those bacteria aren’t going to grow resistance themselves.
I actually can’t imagine *not* using machine learning extensively for anything like this these days. If you’re processing a lot of data, there is likely ML.
We don’t have *any* AI today though, it literally does not exist and everyone would for sure know about it if we did, so AI has not been used and the headline is factually wrong.
Big deal, get ready for the revolution
Indian doctors (heavy breathing): can’t wait to prescribe this for regular fever
Now let’s overprescribe it to people who don’t have bacterial infections /s
Can the AI give us healthcare?
How can they tell the antibiotics are using AI.
Time to push AI to finding effective treatments for more ailments. Hoping for effective treatments, vaccines, and more discoveries.
Does this shit work for cdiff?
This type of thing, and material science, are the 2 main things ML tech is going to be the greatest at. Grinding away at data to narrow down viable, interesting properties from nearly unlimited possibilities. Sorting through immense amounts of data for things is exactly what this technology is perfect for.
The next step of course is even more amazing, but current tech is still good enough to be perfect for this kind of thing.