New superbug-killing antibiotic discovered using AI
bbc.com
bbc.com
> Superbug killing antibiotics with the help of AI
means "A superbug is killing antibiotics, with the help of AI."
> Superbug-killing antibiotics with the help of AI
means "Antibiotics are killing a superbug, with the help of AI."
https://www.merriam-webster.com/dictionary/table#:~:text=tab...
a: to remove (something, such as a parliamentary motion) from consideration indefinitely b British : to place on the agenda
They started with manually testing thousands of drugs, in order to narrow another similarly sized list by one order of magnitude, which was then tested manually. Did they actually save time compared to what it would have taken to test the 6,680 list manually? I guess this needs to go up by one order of magnitude to be really worthwhile?
Think of the poor superbugs!
I think the effort in the testing of the thousands of drugs was to help create the AI model.
I'm no expert on AI, but I get this sense that the exponential improvements that many believe will lead to the singularity may in fact reach an inflection point where the curve flattens out becomes linear or asymptotic, as the rate of improvement is governed by the rate of new information in the real world.
Maybe because they don't pay shit, and can be done by the massive amounts of unskilled labor that exist? Really hard to develop a robot cheap enough for the dexterity needed.
But, even then it's a mistake to think this isn't going to be a massive problem. If everything 'expensive' gets automated then that can lead to a huge pool of labor fighting for low paid jobs that can't actually pay for any assets like houses, education, stocks, etc.
> Most jobs are not well-defined / algorithmic and there is no amount of reading that can prepare you for the embodied, dynamic experience of performing those tasks.
Yea, there is, building an embodied robot and feeding it virtual situations based on real situations. As we get closer and closer to AGI the 'general' functioning of the robot is more and more covered and less and less human intervention is needed.
Here we screened ~7,500 molecules for those that inhibited the growth of A. baumannii in vitro. We trained a neural network with this growth inhibition dataset and performed in silico predictions for structurally new molecules with activity against A. baumannii. Through this approach, we discovered abaucin, an antibacterial compound with narrow-spectrum activity against A. baumannii.
https://www.nature.com/articles/s41589-023-01349-8
Not only were both the dataset they created, and the model they trained on it, specific to one organism, the drug they discovered also only works on that one organism ("narrow spectrum activity against A. baumannii"). If they wanted to discover drugs that work on other organisms, like Staphylococcus aureus and Pseudomonas aeruginosa that the BBC article mentions, they'd have to start all over again.
So, not an approach that looks very practical at this time. Maybe in the future, when the sample efficiency and generalisation ability of neural nets has significantly improved it will be useful in practice.
Study:
If we had a model that could predict the next working antibiotic for MRSA that would be amazing. And you'd probably need it multiple times as MRSA keeps evolving new defenses.
Even narrowing down the list of substances to test by 10x it's amazing.
This looks very practical to me.
We can reasonably expect the bacteria to mutate against the new antibiotic if/once it's used. It's one shifty opponent. This may make the model obsolete, but maybe not - there'd cause to try the model. Actually, it would have been preferable to get more than one result at first...
[EDIT: Then again, would they have another candidate list? This model doesn't do toxicology. The second list was created by using existing proven-safe meds. Do they have another couple thousand materials good to go? If not, they won't be able to run a second time. ]
The 10x better version of it from the future doesn't just appear out of nowhere. We get there by step.
> The AI was then unleashed on a list of 6,680 compounds whose effectiveness was unknown. The results - published in Nature Chemical Biology - showed it took the AI an hour and a half to produce a shortlist.
"published in Nature Chemical Biology" is a link you have to click to see the article in fulltext, which I would encourage you to read if you really want to understand the study. I would link it directly, but there is some site-referrer magic happening that allows the BBC article link to cause Nature publishing group to show the fulltext.
To better understand what was done:
Training set (manually tested): off-patent drugs (2,341 molecules) and synthetic chemicals (5,343 molecules). In particular the synthetic chemicals are likely to have unacceptable side effect profiles. Result of manual screen: 480 molecules capable of inhibiting Acinetobacter growth by 20%. 480/(2341+5343) = 6-7%
Result of AI processing and additional filtering criteria: model applied to "Drug Repurposing Hub" dataset consisting of 6,680 molecules which they claim have demonstrably favorable cytotoxicity profiles and drug-like properties and yielding 3 sets of 240 drugs each. The three sets:
1. 240 drugs identified by the model as having >20% probability of at least 20% growth inhibition of Acinetobacter and structurally _dissimilar_ to those with antibiotic activity in the training set. Manually testing for those capable of more stringent criteria (80% inhibition of growth) yielded 9 drugs.
2. 240 drugs with lowest prediction scores: Manual testing yields no active drugs, providing some basic validation that classifier is functional.
3. 240 drugs with highest prediction scores (without additional filtering criteria based on structural dissimilarity): Manual testing yields 40 drugs capable of 80% inhibition. 40/240 = 16-17%. Yield enrichment: 16%/6% = 260% or a 2.6x improvement compared to naive screen of training set. To be fair, this isn't a direct claim of the paper for good reason: the drugs in their training set and "validation set/drug repurposing hub" are fundamentally different and may have different baseline antibiotic activity across the set.
The process of narrowing down these datasets using the model could be accomplished in hours (their claim) instead of days (my claim). Days is optimistic prediction, requiring high-throughput systems and/or staffing in place to run these screening assays mostly in parallel instead of serially. Also prevents costs associated with further biochemical investigation (cultures, chemical synthesis and assays are not free).
Most direct value of this work is in accelerating drug screening process, reducing cost and developing AI-tractable representations of pharmaceutically-relevant chemical features. Additionally, proof of concept for identifying drugs with appropriate side effect profiles that happen to have antibiotic activity but would not have been identified with existing/common structural analysis approaches, since they are structurally dissimilar to the testing dataset screen. In this case they identified a "CCR2− selective chemokine receptor antagonist" that had antibiotic properties; some googling suggests that this drug class mostly has roles in fibrosis/inflammation regulation and may have roles in autoimmune disorders and those with significant fibrosis as part of the pathology (e.g. cardiovascular disease, liver disease, diabetes). You wouldn't expect most drugs in this class to have any antibiotic properties and many companies would not focus their first efforts on screening such drugs with biochemical assays.
https://www.nature.com/articles/s41589-023-01349-8
I copied it from the Nature article, where I got to from the BBC link. The Nature page has the full text (I'm not logged in). I'm on firefox, is it your browser?
Edit: Oh, wait, I am logged in to Nature. But only on firefox. When I switch to chrome and try navigating to the Nature article by clicking the BBC link, I get a paywall. What exactly do you see on your side?
The top right of the PDF-alike has a link that says "what's this". and it sends me to https://www.springernature.com/gp/researchers/sharedit which seems like a service that allows sharing of links to fulltext. If I go to the url directly, e.g. by copy-pasting the URL or clicking on your link, then I do not get the same behavior. Hence, I deduce that it is probably some referrer magic.
When I click the same BBC link on chrome on android mobile, I don't get the redirect, so probably there is some User Agent stuff too.
>>> We first screened a diverse collection of 7,684 small molecules at 50µM for those that inhibited the growth of A. baumannii ATCC 17978 in Lysogeny Broth (LB) medium (Fig. 1b and Extended Data Fig. 1a). This chemical collection consisted of both off-patent drugs (2,341 mol-ecules) and synthetic chemicals (5,343 molecules) curated from various high-throughput screening sub-libraries at the Broad Institute. Using a conventional hit cutoff of one standard deviation below the mean growth of the entire dataset resulted in 480 molecules being defined as ‘active’ and 7,204 being defined as ‘inactive’ (Supplementary Data 1).Next, this dataset was used to train a binary classifier to predict whether structurally new molecules may display activity against A. baumannii. Briefly, we leveraged a directed message-passing neural network architecture, which translates the graph structure of a mol-ecule into a continuous vector18 (Fig. 1a).This type of model operates by iteratively exchanging informa-tion of local chemistry between adjacent atoms and bonds in a series of ‘message-passing’ steps. Each iteration of message passing propa-gates information about local chemistry across the molecule, thereby allowing the model to build a more holistic representation of the mol-ecule. After a defined number of message-passing steps, the vector representations of various local chemical regions of a molecule are summed into a single continuous vector that captures the complexity of the entire compound. This learned final vector is then supplemented with fixed molecular features computed using RDKit19. A final vector containing both learned and computed features is then used as an input vector for a feed-forward neural network that predicts antibacterial properties. The model was further optimized by using an ensemble of ten classifiers, increasing its robustness. Our final model achieved an area under the precision-recall curve of 0.337±0.088 and an area under the receiver-operating characteristic curve of 0.792±0.042, providing confidence in leveraging the model for predictions in new chemical spaces.
That's an unusual unit.
> Artificial intelligence (AI) and machine learning are often used interchangeably, but machine learning is a subset of the broader category of AI.
and
> Machine learning is a pathway to artificial intelligence. This subcategory of AI uses algorithms to automatically learn insights and recognize patterns from data, applying that learning to make increasingly better decisions.
2. I read about this and see Nvidia up in the news by another 20% today and it makes me think. The old economy "value stocks" in comparison, which are popularly considered to have no exposure to AI, are being dumped by the wayside. But AI advances will benefit those same old economy stocks in many cases. They still possess the necessary distribution to bring them them to market effectively. It's traditional drug companies in this particular example that are best positioned to capitalize on this.
That's not to say that regulating its use isn't a good idea, just that it's really complicated.
From the article:
> Curiously, this experimental antibiotic had no effect on other species of bacteria, and works only on A. baumannii.
This is an oversimplified claim made by the BBC article, but broadly aligns with the research paper claim that the identified drug is specifically not broad-spectrum and not active against e.g. Pseudomonal and Staphylococcal species. The identified drug candidate is unlikely to ever see broad use because of the low activity against many clinically relevant bacteria.
The most common use case for this candidate, should it see approval through the FDA process for this indication, will be for treatment of hospitalized patients who, while waiting for culture results, have failed broad-spectrum regimes and whose culture results demonstrate Acinetobacter infections.
This is currently the situation in India, where the bulk of our antibiotics are manufactured.
ML models like this need training examples and The article talks about manual testing thousands of compounds to gather enough training data. This is where lots of money could have a huge benefit. The more high quality data a model has the better it’s accuracy
This came out of McMaster, though, and I don't know if the Canadian government does funding the same way as we do in the USA.
Still it raises an ugly question, why can discoveries funded by the citizens of a country be turned into the patented property of private companies.
I can name four or five labs on my floor that are currently working on various aspects of antimicrobial resistance, all of which are government funded.
The talking computers goals are "weird", the getting help with solving problems part seems more sane.
Some of it is genuinely antimicrobial stewardship - keeping people off proton pump inhibitors and certain classes of antibiotics.
Fecal transplant is also getting both more viable and better understood (this is what I worked on for my dissertation).
There's also some investigation of alternative therapeutics for C. diff that aren't just vancomycin.
Fortunately a drug that was recently in clinical trials as RBX 2660 [0] was recently approved for use within the US as REBYOTA [1]. It's an improved engineered version of a fecal transplant.
It's actually free even if you don't have insurance although, you have to have two recorded recurrences of Cdiff to be approved. Vancomycin is incredibly expensive, one course WITH insurance was around $385.
I've found a few brands of probiotics work well for maintenance as my recovery continues with some clinically supported probiotics and a diet that avoids trigger foods / too much sugar.
BioGaia [2] - contains L. reuteri which is clinically shown to combat active cdiff
Florastore [3] - creates a biofilm that aids in removing cdiff spores from your gut
BioK + [4] (intended to be used in conjunction with antibiotics to help avoid cdiff / has variants used for patients with feeding tubes to maintain gut health)
At this point, I'm back to being healthy and gaining weight, hoping I don't have serious colon cancer risk factors in the future. Curiously, for those who don't know, (I didn't) when you take Vancomycin you're supposed to suppress calcium intake as much as possible AND avoid any and all trigger foods for gut inflamation.
0 - https://pubmed.ncbi.nlm.nih.gov/36287379/
1 - https://ferringusa.com/?press=ferring-receives-u-s-fda-appro...
2 - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5607411/
Tumeric.