The First Fully AI-Generated Drug Enters Clinical Trials in Human Patients

"The first drug fully generated by artificial intelligence entered clinical trials with human patients this week," reports CNBC: Insilico Medicine, a Hong Kong-based biotech startup with more than $400 million in funding, created the drug, INS018_055, as a treatment for idiopathic pulmonary fibrosis, a chronic disease that causes scarring in the lungs. The condition, which has increased in prevalence in recent decades, currently affects about 100,000 people in the U.S. and can lead to death within two to five years if untreated, according to the National Institutes of Health.

"It is the first fully generative AI drug to reach human clinical trials, and specifically Phase II trials with patients," Alex Zhavoronkov, founder and CEO of Insilico Medicine, told CNBC. "While there are other AI-designed drugs in trials, ours is the first drug with both a novel AI-discovered target and a novel AI-generated design...."

"When this company was launched, we were focused on algorithms — developing the technology that could discover and design new molecules," Zhavoronkov said. "I never imagined in those early days that I would be taking my own AI drugs into clinical trials with patients. But we realized that in order to validate our AI platform, we needed to not only design a new drug for a new target, but bring it into clinical trials to prove that our technology worked."
"The company has two other drugs partially generated by AI in the clinical stage..."

Re:Wonderful

[fahrbot-bot]

Let's keep making things we don't understand and use humans as guinea pigs. In fact, let's hand it over to startups to make it more efficient.

Aspirin has been used for over a hundred years, and willow bark for a few thousand, and scientists still don't know how exactly it works for all the things it works for. (sources: many articles from how does aspirin work [google.com])

Re:

[burtosis]

Moreover aspirin is just the acetyl version of salicylates that are present in most all fruits and vegetables, grains and nuts, and are not from just some rare tree or herb. The methyl version is oil of wintergreen. It’s thought plants use salicylates as a fungicide and insecticide among other functions.

Re:Wonderful

[Miles_O'Toole]

We've actually gone a bit further: https://www.uchicagomedicine.org/forefront/news/how-aspirin-works [uchicagomedicine.org]

Re:

[aldousd666]

I read that UChicago article before, sometime in the early 2000's and I was like 'holy shiz, I thought this was supposed to be a mystery still' and lo, it was not anymore. You just took me back. lol

Re:

[technology_dude]

So this is how it happens? If AI designed drugs becomes the norm and AI continues to advance then ....

Will we let AI design biological drugs? If we don't fully understand the drug, could we spot a "biological timer" that would activate after a certain time period and become cancerous? What if the process of the timer was "encrypted"? Would we consider it to be "junk DNA" and disregard it?

I wonder what AI will do with the planet after it inherits it?

Re:

[kvezach]

Most drugs are small molecules, which kind of acts like regularization. It wouldn't be easy for an AI with the wrong incentives - or a company with the wrong incentives running an AI - to hide a sophisticated function like a timer in a small-molecule drug.

This wouldn't hold for large things like monoclonal antibodies, but I guess you could make AIs that would try to detect anything strange. If you're making an AI with the purpose of creating drugs, it has to recognize what makes for a good drug, and if t

Re:

[technology_dude]

Tidbit #1
There are several biological drugs on the market. One of them gave my co-worker leukemia. She beat that only to have another biological drug used to treat graft vs host disease to cause her to have a stroke and she died.

Tidbit #2
A recent thought experiment resulted in a possible scenario where an AI weapon took out it's human overseer because the human was interfering with it's mission.

Those two tidbits together shows it is not a far stretch for a bad guy to design an AI to develop a dru

Re:

[Anonymous Coward]

It's called science. This is how progress is made

Re:

[quonset]

Let's keep making things we don't understand and use humans as guinea pigs. In fact, let's hand it over to startups to make it more efficient.

Humans should be guine pigs because we're the ones who will be using the drugs. No need to involve any other animal.

Re:Wonderful

[sound+vision]

The traditional approach to developing pharmaceuticals isn't any worse than AI. You have no idea what the compound does until you stick it in somebody. Whether the molecule was dreamt up by a human or a machine has no bearing on that.

At least they run rigorous trials on these compounds before spreading them all through environment and the food supply. You can't say the same for compounds developed for consumer goods.

Re:

[backslashdot]

There's no drug we can possibly "fully understand."

Patents ???

[DDumitru]

It might not be possible to patent an "AI Invented" Invention. The Copyright Office has already chimed in to preclude copyrights. The legal landscape could get very complicated as Inventions require "Inventors" which are generally "human persons". You will note that patents list people and not companies as inventors.

Re:

[TheRealMindChild]

Who gives a shit? Why would it need to be patented?

Re:

[Can'tNot]

To make as much money as possible. Here, let me quote the relevant bit from the summary: "Insilico Medicine, a Hong Kong-based biotech startup"

Re:

[Ellis D. Tripp]

Because capitalism.

Re:

[TheRealMindChild]

Without it being patented/patentable, then everyone can have at it and let the market sort it out through competition. That sounds a lot more like capitalism to me

Re:

[xlsior]

1) Even if an AI spits out a list of potential drug candidates, the company may list the person who actually fabricated the drug and verified that it worked as the inventor?

2) Even if the drug composition itself could not be patented, they can still patent their manufacturing/production method to produce the drug, and use that to keep the competition at bay.

Re:

[xlsior]

Aside from the "a computer did it" aspect: Since modern AI's basically just regurgitate and recombine existing knowledge you could argue that every single thing that an AI spits out could be considered "obvious" and would therefore not be eligible to be patented on those grounds alone.

Re:

[micheas]

Aside from the "a computer did it" aspect: Since modern AI's basically just regurgitate and recombine existing knowledge you could argue that every single thing that an AI spits out could be considered "obvious" and would therefore not be eligible to be patented on those grounds alone.

It isn't so much regurgitate and recombine as take the next logical step. Which I would think makes your point even stronger.

Re:

[real_nickname]

That's because AI is a software, you don't credit a camera firmware as a photographer. The "AI Invented" thing is just a PR stunt.

The real question

[marcle]

How much are they gonna charge for it?

Re:The real question

[jacks smirking reven]

If you have idiopathic pulmonary fibrosis (particularly in the US) it's an important question.

Re:The real question

[schematix]

Same as everything else.... whatever the market will bear.

Re:

[narcc]

That doesn't really work when it comes to people's health.

Re:

[Can'tNot]

Well it doesn't keep people healthy or financially solvent, but as long as those things aren't how you measure success: it works great.

We'll see how it goes

[Todd Knarr]

The Phase I results were basically "it didn't hurt or kill anyone". I notice they didn't say anything about it having any actual positive effect. Phase I is primarily about safety, but the rule of thumb is still that your efficacy results don't improve as you move through clinical trial phases. If they aren't shouting positive Phase I results from the rooftops, that suggests they weren't any better than placebo which doesn't bode well for Phase II.

People on the drug unworried about AI

[Babel-17]

So a complete success?

They're doing an IPO.

[Truth_Quark]

Not suggesting that the good news story is exaggerated at all. But insilico medicine has filed for a Hong Kong IPO [forbes.com], so either this sort of publicity is timely, or they're doing the IPO on the strength of this drug going to trial ...

Which could mean that they want some pocket money to rapidly expand their operation, or it could mean that they want to spread the risk of the trial being a failure. (Or some combination of the two).