I did this at home with a 5090 on a different target a few weeks ago. Cannot generate 318,000 designs but Claude chose the nearly same tooling, except I used BindCraft2 which is better for multi target designs in the beginning. This isn't that special, its just doing tool calls. I got 100 or so trajectories running 24hrs just for fun.
"Throughout the process, Claude generated 318,484 protein designs and selected 180 of them for testing. Our team added 31 more, with Claude’s help."
"lab-tested all 211 designs to see if they would bind to the toxins they were designed against, and 17 designs met our criteria for a binder. "
Alright how many of those 17 were 100% Claude, I know mine were. These mini-proteins the tooling I'm told is pretty good for these. I'm worried that Claude might just choose the same tooling for everyone if this is the path we go down and designs stagnate?
daemonk 5 hours ago [-]
Claude will just choose similar toolings as everyone. Orchestration is a really nice tool to have and this can definitely empower non-technical bench biologists. But the bottleneck isn't using the tools or even intelligence in this field. It's data generation and ontology. There needs to be more hardware automation and good data/record keeping in biology. With the exception of high-throughput sequencing and some other more niche quantitative methods, I feel like molecular biology is still at similar data throughput of Mendel's pea experiments.
Frost1x 23 hours ago [-]
There’s certainly value in using LLM to search unknown spaces for research of knowledge captured in human text forms, but having done a lot of modeling and simulation in applied sciences over my career, I often wonder for some of these cases how much is just having people interested in various computational sciences, modeling and simulation, and a massive pool of compute to deploy things and approaches to rapidly wire things together on the cheap, sort of like this case.
automationwiz 22 hours ago [-]
It was pretty grand not having to string together the bio-informatics and protein design pipeline and let Claude handle that and the environment. A lot of these tools in the past are fragmented and sometimes frozen to certain versions of a language which meant debugging. I didn't have to do at all; if there was env issues or GPU incompatibility Claude ran tests before execution and figured it all out for me. Truly the most boring stuff of this it fixed which would probably have made me give up if I was on my own.
conorh 1 days ago [-]
I hope this produces some useful anti-venom that is cheaper and widely available. My kid was stung by a small bark scorpion on his thumb here in Arizona and I would not wish that experience on anyone. For hours afterwards he was in excruciating screaming pain, and he even lost his vision for while. That lasted about 24 hours and then it took weeks for him to recover feeling in his hand. Nothing seemed to help with the pain and there was very little we could do for him. I would love to have had a more available anti-venom, but apparently now it is so expensive they only keep it around for the very worst cases.
zardo 23 hours ago [-]
I guess if you want good health care at an affordable cost you should have gone to Mexico where the anti-venom that is $10,000 a vial in the US is sold (by the same company) for $100 a vial.
gradus_ad 23 hours ago [-]
It's only $100 in Mexico because the company can charge $10000 in the US. Not by some inherent virtue of the Mexican system. If the US forced costs down they'd rise everywhere else.
The US subsidizes the rest of the world in Medicine, Defense, Technology, etc...
torginus 39 minutes ago [-]
Thanks to the magic of Google, I've learned that Mexican scorpion antivenom was in fact developed in Mexico, by a Mexican company, and now you did too.
Is the 10K price in the US due to ongoing costs, recouping R&D, paying for marketing? Because if it's not a real ongoing cost then it's just "wha'tcha got?" pricing.
I doubt they sell it so cheap in Mexico to be charitable. Which suggests they're just soaking the US market.
elromulous 23 hours ago [-]
This could be simply fixed by the US allowing the re-importation of drugs. Without it, we end up with monopolistic pricing.
jknoepfler 23 hours ago [-]
note that this would be extremely bad for the poorer country that is currently able to purchase medicine, and would probably results in net worse health outcomes globally.
the free market is just not the correct way to set drug prices, and never has been.
jacquesm 23 hours ago [-]
I wonder how you justify the cost of insulin.
analognoise 23 hours ago [-]
The lies Americans tell themselves to support their own exploitation will always amaze me.
Viliam1234 3 hours ago [-]
"The reason we don't provide free healthcare to our citizens is because we have decided to provide 99% discounts for the rest of the planet instead."
Next they will tell you that they have so many homeless people because they are sending all their houses to Mexico.
jknoepfler 23 hours ago [-]
I don't think it's rational to believe that big pharma is giving charity to Mexico. It's $100 a vial in Mexico because that's the optimal price to sell it in Mexico to maximize profit, not because it's subsidized by unregulated price-gouging in the United States.
gradus_ad 23 hours ago [-]
Does Mexico have a free market for medicine?
jknoepfler 9 hours ago [-]
Why would that matter? The drug company will price their product to maximize profit in whatever regulatory regime they're selling in. (I have no idea).
snsifnrj 17 hours ago [-]
Oh hey look, a brain-dead cultist!
Can’t wait for dang to come in and say it’s OK to be racist.
kevin_thibedeau 1 days ago [-]
> but apparently now it is so expensive
They saw Shkreli hike prices 100x and decided that was the business model for them. Same thing has happened with rabies.
darth_avocado 1 days ago [-]
A single shot of rabies costs about ~$3 in Asian countries where street dogs are common. It’s crazy how expensive post exposure rabies treatment is in the US.
skeledrew 23 hours ago [-]
In Asian countries, people needing treatment is an opportunity to improve lives. In the US, people needing treatment is an opportunity for someone to make bank.
kenhwang 23 hours ago [-]
Pretty sure most of the post-exposure (PEP) cost is from the immunoglobulin, which is also why antivenom is often so expensive because it is also often immunoglobulin based. A synthetic alternative to immunoglobulin should bring the price down and make it easier to store/transport.
Pre-exposure (PrEP) for rabies (a vaccine) is more expensive than it should be due to lack of economics of scale and the typical American medical pricing inflation. That's not unique to rabies, it's just how vaccines are allowed to be priced in the US.
My understanding is poorer rabies prone countries do not even attempt PEP for rabies due to the cost, they really rely on broad PrEP immunity.
PEP is always going to be more expensive than PrEP, because transplanting a vaccinated animal's immune response proteins is a much harder thing to do than getting your body to make it's own.
trashface 16 hours ago [-]
Vision involvement should have qualified your son for antivenom. Sounds like good ol' US healthcare fuck up, again. Every time.
LoganDark 1 days ago [-]
Unfortunately, available does not mean cheap these days. Parents (good ones anyway) would go into a decade of debt to prevent their child from having an experience like you just described. Drug giants see that as a business opportunity.
dreamcompiler 1 days ago [-]
This might be a job for fentanyl. Carefully monitored in a hospital setting of course.
amelius 1 days ago [-]
Leave it to the internet to come up with a new application for a recreational drug.
Had to get fentanyl during a gall bladder issue, only thing that stopped the screaming
dreamcompiler 22 hours ago [-]
My paramedic colleagues were using fentanyl decades before you heard of it in a recreational context.
vhiremath4 1 days ago [-]
Can someone explain whether design of protein binders is an important rate-limiting step in drug creation (of which antivenom is a very small market)? Or AI-driven drug discovery in general? It feels like something that will definitely help the drug market but is nowhere near the rate-limiting step, but that's from my own researcher as someone without a background in it.
My understanding is that FDA trials (specifically phase 2 + 3) are the death valley for most drugs and finding a biomarker is something mostly done through a mix of gene bank scraping + looking at old biomarker databanks. The majority of the cost and time is in failed drugs that do not pass FDA trials (which average 6-7 years end-to-end).
Note: I am asking this in earnest. I am not trying to ask a leading question and saying it is not important at all.
devindotcom 1 days ago [-]
My understanding is it is not - this is something we have models for, and there was a flurry of machine learning based drug discovery companies about 8-10 years ago. But they ran into the testing bottleneck real fast because they would spit out thousands of candidates for binders and so on that all needed to be tested in vitro before they could even start the real drug dev circuit. Furthermore the drugs need to be manufacturable and that's not a trivial problem.
It's nice to (as the blog post suggests) give the opportunity to non medical experts, say in this case a snake/venom research center, to pursue a problem and produce some realistic candidates that they could pay to test. I do think it's good to democratize what parts of the process can be, but it's more of a resource prioritization problem than a molecular design one.
entee 24 hours ago [-]
Have a comment below on this, but generally things don't work out in the clinic because we didn't understand the biology adequately, and not because we didn't have a molecule.
A better binder only helps if you know what to bind to and that if you bind to it, you'll have the desired effect.
This is why the industry has crowded around a very limited set of "validated" targets. It's because we know if we do something with them, they will work. Obviously we're likely missing a lot of great ideas though.
AI might help find some of those ideas, but we won't know until someone tests them in a human. Essentially, without more extremely expensive and difficult data to acquire, it's not likely to be that much better at picking winners.
I'm in the sector and I believe strongly in AI helping (I've bet my entire career on it). It will get there eventually, but it will take much longer than people seem to think at the big labs.
Parent comment reached out asking for examples so thought I'd post my answer here:
We thought clearing amyloid beta plaques would slow alzheimer’s disease. We now have antibodies that do clear the plaques quite effectively, however they barely show any clinical benefit. Some trials failed entirely, others cleared a statistical bar that clinicians agree would be hard to detect by someone just living with the person. Essentially all patients given this drug still decline at more or less the same rate they declined before getting the drug. Definitely not the “we took away the cause of the disease, so at least they’ll stay about the same as they are today” effect we were hoping for.
To find this out, many trials costing (hard to get the specific number) but hundreds of millions of dollars had to be run. And this was easily the consensus target for solving Alzheimer’s.
If AI magically comes up with a new idea (Eliminate Protein X and Alzheimer’s will stop in its tracks), it will take us that long and that much money to know if it was right.
And before that we’ll have to run a huge number of lab experiments, likely involving developing new model systems (engineered mice usually, but not always) to test the concept before a human gets the drug.
So even assuming an AI model today is a “Genius” level talent, and that the target we need to go after is really already in some pre-existing paper or dataset, it will take a huge amount of investment to know.
And again, likely we need more data about how Alzheimer's works, because our best ideas haven’t worked the way we expected. Maybe AI can help us design ways to get that data, or ideas for what to gather. But it’s not going to magically come up with an answer where we simply do not know.
There’s a huge cardiology trial using Lp(a) as a target that just failed because it turns out that lowering cholesterol using the drug doesn’t actually reduce cardiac events. This was the most genetically validated open target in cardiology, and it turns out to not work how we thought. Hundreds of millions of dollars lost. (https://www.cnbc.com/2026/09/08/novartis-cholesterol-setback...)
Take another unrelated example: If AI told me, “I’ve done a bunch of geology, and I know there’s a huge vat of oil under the ground over there”. You’re gonna need to drill to find out. And if you give the AI all the information you have about geology and it says, “Maybe there’s oil under the ground over there, but I can’t be sure unless you take these rock samples”. Then someone needs to go out and get the rock samples, grind them up, stick it in a machine, get the data.
All those things take time and money. Ideas are cheap, measurements cost time and money. In drug discovery, we’re often against the frontier of what we know. Past the frontier you need new data. In computational fields (coding, video games, maybe in some ways mathematics) you can convert compute into data. In biology and the physical world, you have to do measurements. Compute can’t help you at the edge of data in the physical world.
vhiremath4 17 hours ago [-]
Extremely helpful and thank you for relaying back here!
eig 1 days ago [-]
While Clinical Trials are the most expensive step and have a 90% failure rate, it is not the *source* of the failures, it's just where failures are revealed.
Much of whether or not a candidate drug will succeed was determined all the way back in the design/discovery stage. If we can design better drugs, then we would enjoy a higher probability of success in the later clinical trials.
The single biggest improvement to pharma drug development would be from increasing the probability of success in clinical trials.
vhiremath4 1 days ago [-]
Is it fair to say the source of the failures is we simply do not understand the complex relationship between the drug design (chemistry, cell therapy, etc.) and the biology (the complete system of the living thing). If we could understand that relationship, we could design better drugs? I know this is very hand-wavy. Probably so much so as to not be a useful thing to say, but I'm just trying to understand.
slashdave 1 days ago [-]
Not necessarily. It is often due to a lack of choices (in lead candidates).
EA-3167 1 days ago [-]
Correct. The on-target effects of antivenin can be well-modeled along with a known subset of off-target effects, but the latter is far too complex and unknown to model completely.
Trials are the acceptable middle ground that uses animal and human models to explore those unknowns.
entee 24 hours ago [-]
Some trials fail because we didn't have a "better" drug for a given target, many fail because we don't know the right biology. Biology and knowing precisely how to modulate the right thing is more gating than "design a better compound".
Better for what is the question and while sometimes the answer is "solubility, metabolism, off target toxicity", more often it's "you misunderstood something about the biology of this disease".
Maybe you needed more potency than your assays said you needed.
Maybe you missed something else the target does.
Maybe you had the right effect on the target but the body found a way to compensate (in an undesirable way).
Maybe the target looks like it matters in all sorts of model systems, but in real life it doesn't matter at all!
Many such cases. AI might marginally help in these scenarios, but doesn't look like it will eliminate the problem. AI can't help us with what we fundamentally don't know and have zero data for, or for which data is extremely hard to get.
Otterly99 14 hours ago [-]
Disclaimer: I did not work in protein design but in biophysics. However some of my collaborators did.
It definitely is a small help. Although the lab I was working with had supercalculators to generate microtubule-binding proteins, They were kind of limited in how many they could simulate and try, so being able to go through 318K molecules is useful. Our lab was then tasked with doing the testing (I designed the automated assay) and this part is going on right now, I expect it to last around a year (small lab, many tests to run).
Once potential medical targets have been found, the actual work takes several years as you have mentionned as you would switch to cell cultures, then animal models, then human trials. That matches the 6-7 years timeline you mentionned.
So overall, you can crunch a ~6-months period into 1 month, which is definitely a time saver.
refurb 12 hours ago [-]
As a former drug discovery scientist I agree with the “small help” conclusion.
Find “potential leads” is something we did plenty of with computational chemistry. We’d get leads that we a “slam dunk” and they’d fail horribly in clinical trials.
Finding leads is easy, finding leads that reliably pan out in the lab is not
dooglius 1 days ago [-]
The impression I have, which admittedly is purely as a layman reading articles, is that by far the rate-limiting factor in all drug creation is the cost of doing clinical trials in humans.
1 days ago [-]
ray__ 23 hours ago [-]
Protein binders (as in, small proteins like the one in the article that themselves bind other proteins) aren't a rate-limiting step, but also are fairly restricted in their applications as therapeutics. Designing small-molecule ligands is much more challenging, and AI/ML design tools for small molecules are much less effective for a number of reasons, including 1) existing training data are sparse the search space for small molecules is arguably much larger and more complex than for proteins 2) the evolutionary data that powers protein structure prediction and design don't exist for small molecules. So it is still non-trivial to design a small-molecule binder for a protein.
It's true that clinical trials are a big cost center for drug development, but there is an argument for mitigating failure here earlier in the process through better technology development. Some of this will come from AI and some of it from better experimental methods for probing biology. One of my favorite efforts in this area is OpenADMET, which combines ML and experimental methods for addressing tox-related issues in drug discovery, which are common causes for failure in the later (more expensive) clinical trial stages:
Your impression is correct. Clinical trials are the rate limiting step and the primary source of cost and risk.
slashdave 1 days ago [-]
There would be less failures if problematic therapies were filtered out before the clinical stage.
cyberax 23 hours ago [-]
Both "yes" and "no". AI has been doing great in protein structure prediction.
It's a task that is well suited for the current generation of AIs, (most) proteins consist of a constrained "alphabet" of components, and the alphabet then forms recognizable shapes ("motifs"). So working backwards and generating a protein that binds with another protein is the logical next step.
The problem is in the efficacy. Monoclonal antibodies work surprisingly poorly as antivenoms. They are too specific, and real venom is a mix of multiple kinds of proteins with a lot of variation. Which makes sense, as venoms are under high selective pressure, so there's going to be a lot of genetic variability within a single species (never mind _across_ species).
Snake antivenoms are a major underserved area with a decent amount of research. But so far the century-old method of producing antivenoms (injecting horses and then collecting their blood plasma) is the best one.
sidewndr46 24 hours ago [-]
my general question with any drug isn't its effectiveness. It is what are the side effects? Unless somehow AI is solving this without clinical trials, there isn't much of note here
pazimzadeh 21 hours ago [-]
it is a rate-limiting step.
previously you'd inject a rabbit or other small mammal with what wanted antibodies to bind against, plus an adjuvant compound
then you'd need to screen the antibodies and figure out what they do. some antibodies will bind to your target but not neutralize it, so have to rule those out.
then I think you'd have to do amino acid sequencing on it I think and then clone the sequence in some B-cells to produce it.
there's probably a way better approach than what I just said but it's usually still going to start with stabbing an animal with your target.
daemonk 1 days ago [-]
It orchestrated available open source tools to de novo generate and assess the designs. It is still really impressive that it has enough theory to use the tools correctly, but it isn't generating the residues from scratch within the model itself.
I got Opus5 to write me a enzyme design pipeline on AWS last month using available tools and databases. It works fine. It's not as handsoff as you might think when you start looking at more complex proteins with multiple binding pockets and sequential catalytic process within multimeric structures. The bottleneck is still actually testing the candidates out in the lab.
MattGrommes 1 days ago [-]
This sounds awesome but I wish these would be published in a peer-reviewed journal or something of that nature. I can't believe these press releases they put out, even though I'm sure it's an achievement.
giancarlostoro 1 days ago [-]
They're trying to catch up to OpenAI which has been doing these a lot, but I agree, can we actually get them peer reviewed as a hard requirement? I think OpenAI would have some looked over by experts before posting their blog posts.
enraged_camel 1 days ago [-]
>> They're trying to catch up to OpenAI which has been doing these a lot
Sorry, what? OpenAI is "ahead" in math. Anthropic is "ahead" in life sciences, and even have their own wet lab.
giancarlostoro 1 days ago [-]
How do you determine which one's ahead in what? OpenAI isn't just known for their models being used in mathematical discoveries, they've also been used in other areas as well? Anthropic on the other hand has not had anywhere near the amount of news coverage as OpenAI gets for their discoveries / efforts.
If I were at Anthropic, I'd be making sure the higher ups highlight discoveries made using the models they make, because OpenAI is definitely highlighting those.
enraged_camel 1 days ago [-]
>> How do you determine which one's ahead in what?
We are talking about discoveries the labs have publicly showcased.
1 days ago [-]
aaron695 1 days ago [-]
[dead]
dev_hugepages 1 days ago [-]
This is somewhat funny given that not that long ago they were doing an article instigating fears of what Claude could do, and one of them was a researchers screening venom peptides along other things
Perhaps he could have worked on a similar subject? Who knows.
xvector 23 hours ago [-]
It doesn't take a genius to realize that you should let anonymous randoms do bio "research" on your API. I'm glad they prevent this unless you are partnered with a university, tbh.
AnodicElegy 1 days ago [-]
From the technical report:
"Our prompt specified the targets, the design sets and their sizes, the rules for each design, the protein design and prediction tools, the scores and ranking formulas, and the novelty and diversity rules."
I think most scientists would consider LLM to be an orchestration tool in cases like this. LLMs do work great for automating this type of drug discovery workflow! You definitely don't need the latest and greatest model, either, as long as you're careful in your specification. But attributing the designs to the tool (Claude), not the scientists who came up with the workflow, is distasteful and misleading.
throwaway_20161 1 days ago [-]
So it’s ok for Anthropic to make new biologics … but it’s users cannot because it might end humanity.
Let alone the fact this doesn’t even appear to be peer reviewed either from Anthropic.
runj__ 1 days ago [-]
Yes, they _hope_ they can verify that the outcome of giving someone they trust a gun will not end in the end of the world.
I wouldn't trust you with a gun of mine... I don't know you.
1 hours ago [-]
throwaway_20161 1 days ago [-]
And what makes you trust them, with the “gun”?
red75prime 1 days ago [-]
It's less likely to have an omnicidal maniac in a group of 100 people than in a group of a few billion people.
yunnpp 1 days ago [-]
That assumes you have no prior knowledge of either group.
red75prime 19 hours ago [-]
Which prior knowledge do we have? A group of engineers managed by public figures that show no signs of omnicidality. A large group of people of all walks, which is very hard to vet.
xvector 23 hours ago [-]
No, this is just basic probability.
mapmeld 23 hours ago [-]
In 2025 there was a similar paper using RFdiffusion to design snake antivenom. That team went as far as validating the best antivenom candidates protected mice: https://www.nature.com/articles/s41586-024-08393-x
bayarearefugee 1 days ago [-]
Neat, but doesn't change anything about the fact that even with this protein design headstart nobody is going to pay for the human trials or the cost to produce and distribute the possible antivenom.
In practical terms this is like Claude designing you a new home. Cool, you now have blueprints for a new home. You can print it out and hang it on the wall of your studio apartment.
sharts 21 hours ago [-]
Why not have claude fix our political systems?
sligbad 24 hours ago [-]
I hate how the role of the human scientists involved is so diminished, with credit being given to "Claude," which is at best a family of products.
downrightmike 23 hours ago [-]
Does it cost more than some generic benedril? Otherwise no biggie
waffletower 1 days ago [-]
I don't see scientists reacting to this with the same affront as Terrance Tao does for frontier LLM mathematical proof dumping. Just imagine the Salk Institute declaring "we are on a war footing" after learning of such LLM aided discoveries.
paimapi 1 days ago [-]
[flagged]
squidmaster 1 days ago [-]
[dead]
Catloafdev 1 days ago [-]
Is the CSS on the page all messed up for anyone else or just me?
"Throughout the process, Claude generated 318,484 protein designs and selected 180 of them for testing. Our team added 31 more, with Claude’s help."
"lab-tested all 211 designs to see if they would bind to the toxins they were designed against, and 17 designs met our criteria for a binder. "
Alright how many of those 17 were 100% Claude, I know mine were. These mini-proteins the tooling I'm told is pretty good for these. I'm worried that Claude might just choose the same tooling for everyone if this is the path we go down and designs stagnate?
The US subsidizes the rest of the world in Medicine, Defense, Technology, etc...
https://synapse.patsnap.com/article/what-is-serum-antiscorpi...
I doubt they sell it so cheap in Mexico to be charitable. Which suggests they're just soaking the US market.
the free market is just not the correct way to set drug prices, and never has been.
Next they will tell you that they have so many homeless people because they are sending all their houses to Mexico.
Can’t wait for dang to come in and say it’s OK to be racist.
They saw Shkreli hike prices 100x and decided that was the business model for them. Same thing has happened with rabies.
Pre-exposure (PrEP) for rabies (a vaccine) is more expensive than it should be due to lack of economics of scale and the typical American medical pricing inflation. That's not unique to rabies, it's just how vaccines are allowed to be priced in the US.
My understanding is poorer rabies prone countries do not even attempt PEP for rabies due to the cost, they really rely on broad PrEP immunity.
PEP is always going to be more expensive than PrEP, because transplanting a vaccinated animal's immune response proteins is a much harder thing to do than getting your body to make it's own.
Approved for medical use in 1968
My understanding is that FDA trials (specifically phase 2 + 3) are the death valley for most drugs and finding a biomarker is something mostly done through a mix of gene bank scraping + looking at old biomarker databanks. The majority of the cost and time is in failed drugs that do not pass FDA trials (which average 6-7 years end-to-end).
Note: I am asking this in earnest. I am not trying to ask a leading question and saying it is not important at all.
It's nice to (as the blog post suggests) give the opportunity to non medical experts, say in this case a snake/venom research center, to pursue a problem and produce some realistic candidates that they could pay to test. I do think it's good to democratize what parts of the process can be, but it's more of a resource prioritization problem than a molecular design one.
A better binder only helps if you know what to bind to and that if you bind to it, you'll have the desired effect.
This is why the industry has crowded around a very limited set of "validated" targets. It's because we know if we do something with them, they will work. Obviously we're likely missing a lot of great ideas though.
AI might help find some of those ideas, but we won't know until someone tests them in a human. Essentially, without more extremely expensive and difficult data to acquire, it's not likely to be that much better at picking winners.
I'm in the sector and I believe strongly in AI helping (I've bet my entire career on it). It will get there eventually, but it will take much longer than people seem to think at the big labs.
On crowding in big pharma: https://www.lek.com/insights/life-sciences-pharma/biopharma-...
We thought clearing amyloid beta plaques would slow alzheimer’s disease. We now have antibodies that do clear the plaques quite effectively, however they barely show any clinical benefit. Some trials failed entirely, others cleared a statistical bar that clinicians agree would be hard to detect by someone just living with the person. Essentially all patients given this drug still decline at more or less the same rate they declined before getting the drug. Definitely not the “we took away the cause of the disease, so at least they’ll stay about the same as they are today” effect we were hoping for.
To find this out, many trials costing (hard to get the specific number) but hundreds of millions of dollars had to be run. And this was easily the consensus target for solving Alzheimer’s.
If AI magically comes up with a new idea (Eliminate Protein X and Alzheimer’s will stop in its tracks), it will take us that long and that much money to know if it was right.
And before that we’ll have to run a huge number of lab experiments, likely involving developing new model systems (engineered mice usually, but not always) to test the concept before a human gets the drug.
So even assuming an AI model today is a “Genius” level talent, and that the target we need to go after is really already in some pre-existing paper or dataset, it will take a huge amount of investment to know.
And again, likely we need more data about how Alzheimer's works, because our best ideas haven’t worked the way we expected. Maybe AI can help us design ways to get that data, or ideas for what to gather. But it’s not going to magically come up with an answer where we simply do not know.
There’s a huge cardiology trial using Lp(a) as a target that just failed because it turns out that lowering cholesterol using the drug doesn’t actually reduce cardiac events. This was the most genetically validated open target in cardiology, and it turns out to not work how we thought. Hundreds of millions of dollars lost. (https://www.cnbc.com/2026/09/08/novartis-cholesterol-setback...)
Take another unrelated example: If AI told me, “I’ve done a bunch of geology, and I know there’s a huge vat of oil under the ground over there”. You’re gonna need to drill to find out. And if you give the AI all the information you have about geology and it says, “Maybe there’s oil under the ground over there, but I can’t be sure unless you take these rock samples”. Then someone needs to go out and get the rock samples, grind them up, stick it in a machine, get the data.
All those things take time and money. Ideas are cheap, measurements cost time and money. In drug discovery, we’re often against the frontier of what we know. Past the frontier you need new data. In computational fields (coding, video games, maybe in some ways mathematics) you can convert compute into data. In biology and the physical world, you have to do measurements. Compute can’t help you at the edge of data in the physical world.
Much of whether or not a candidate drug will succeed was determined all the way back in the design/discovery stage. If we can design better drugs, then we would enjoy a higher probability of success in the later clinical trials.
The single biggest improvement to pharma drug development would be from increasing the probability of success in clinical trials.
Trials are the acceptable middle ground that uses animal and human models to explore those unknowns.
Better for what is the question and while sometimes the answer is "solubility, metabolism, off target toxicity", more often it's "you misunderstood something about the biology of this disease".
Maybe you needed more potency than your assays said you needed.
Maybe you missed something else the target does.
Maybe you had the right effect on the target but the body found a way to compensate (in an undesirable way).
Maybe the target looks like it matters in all sorts of model systems, but in real life it doesn't matter at all!
Many such cases. AI might marginally help in these scenarios, but doesn't look like it will eliminate the problem. AI can't help us with what we fundamentally don't know and have zero data for, or for which data is extremely hard to get.
It definitely is a small help. Although the lab I was working with had supercalculators to generate microtubule-binding proteins, They were kind of limited in how many they could simulate and try, so being able to go through 318K molecules is useful. Our lab was then tasked with doing the testing (I designed the automated assay) and this part is going on right now, I expect it to last around a year (small lab, many tests to run).
Once potential medical targets have been found, the actual work takes several years as you have mentionned as you would switch to cell cultures, then animal models, then human trials. That matches the 6-7 years timeline you mentionned.
So overall, you can crunch a ~6-months period into 1 month, which is definitely a time saver.
Find “potential leads” is something we did plenty of with computational chemistry. We’d get leads that we a “slam dunk” and they’d fail horribly in clinical trials.
Finding leads is easy, finding leads that reliably pan out in the lab is not
It's true that clinical trials are a big cost center for drug development, but there is an argument for mitigating failure here earlier in the process through better technology development. Some of this will come from AI and some of it from better experimental methods for probing biology. One of my favorite efforts in this area is OpenADMET, which combines ML and experimental methods for addressing tox-related issues in drug discovery, which are common causes for failure in the later (more expensive) clinical trial stages:
https://www.nature.com/articles/s41467-026-73410-8 https://openadmet.org
It's a task that is well suited for the current generation of AIs, (most) proteins consist of a constrained "alphabet" of components, and the alphabet then forms recognizable shapes ("motifs"). So working backwards and generating a protein that binds with another protein is the logical next step.
The problem is in the efficacy. Monoclonal antibodies work surprisingly poorly as antivenoms. They are too specific, and real venom is a mix of multiple kinds of proteins with a lot of variation. Which makes sense, as venoms are under high selective pressure, so there's going to be a lot of genetic variability within a single species (never mind _across_ species).
Snake antivenoms are a major underserved area with a decent amount of research. But so far the century-old method of producing antivenoms (injecting horses and then collecting their blood plasma) is the best one.
previously you'd inject a rabbit or other small mammal with what wanted antibodies to bind against, plus an adjuvant compound
then you'd need to screen the antibodies and figure out what they do. some antibodies will bind to your target but not neutralize it, so have to rule those out.
then I think you'd have to do amino acid sequencing on it I think and then clone the sequence in some B-cells to produce it.
there's probably a way better approach than what I just said but it's usually still going to start with stabbing an animal with your target.
I got Opus5 to write me a enzyme design pipeline on AWS last month using available tools and databases. It works fine. It's not as handsoff as you might think when you start looking at more complex proteins with multiple binding pockets and sequential catalytic process within multimeric structures. The bottleneck is still actually testing the candidates out in the lab.
Sorry, what? OpenAI is "ahead" in math. Anthropic is "ahead" in life sciences, and even have their own wet lab.
If I were at Anthropic, I'd be making sure the higher ups highlight discoveries made using the models they make, because OpenAI is definitely highlighting those.
We are talking about discoveries the labs have publicly showcased.
"Our prompt specified the targets, the design sets and their sizes, the rules for each design, the protein design and prediction tools, the scores and ranking formulas, and the novelty and diversity rules."
I think most scientists would consider LLM to be an orchestration tool in cases like this. LLMs do work great for automating this type of drug discovery workflow! You definitely don't need the latest and greatest model, either, as long as you're careful in your specification. But attributing the designs to the tool (Claude), not the scientists who came up with the workflow, is distasteful and misleading.
Let alone the fact this doesn’t even appear to be peer reviewed either from Anthropic.
I wouldn't trust you with a gun of mine... I don't know you.
In practical terms this is like Claude designing you a new home. Cool, you now have blueprints for a new home. You can print it out and hang it on the wall of your studio apartment.
https://i.imgur.com/fRVxjzq.png