In just a few days, AI has been able to break through 10 60-dimensional chemical universes and the p

  It's the first time in the history of the world.

In 2021, five years ago, when DeepMind founder Demis Hassabis announced the creation of a brand new AI pharmaceutical company, Isomorphic Labs, the entire medical community was filled with curiosity and skepticism: "Can AI really make drugs?"

Today marks the end of five years.

Max Jaderberg, president of Isomorphic Labs, handed out a five-year report:AI has not only unlocked completely new treatments for a wide range of diseases, but has also completely paid for drug discovery.

In the past, conventional pharmaceutical companies had to complete 10 60-quarter chemical space searches, which took years, and now AI does it directly in a matter of days.

After rigorous wet-testing, these new molecules, which AI "forged in a vacuum", are not only remarkable for their medicinal properties, but also reveal completely new biological mechanisms, opening up pockets of unknown targets, and making a big step towards clinical trials.

Isomorphic Labs is declaring to all humanity:An industrial revolution that has completely disrupted traditional pharmaceutical laws has officially broken out.

To understand how terrifying Isomorphic Labs' response is, we must first understand how desperate "astronomical barriers" are facing traditional drug development.

In the pharmacophysical chemistry space, the number of compounds that can become small molecule drugs is approximately60 times 10 is 10 squared.The total number of atoms in our entire solar system combined is only 57 times 10.

It can be said that finding a perfect new drug molecule is more difficult than finding a specific sand in the entire universe.

What is the traditional pharmaceutical industry doing in the face of such a shaky "chemical universe"?

The answer is:High throughput screening (HTS).

Pharmaceutical companies spend months or even years testing a huge machine stream containing a molecular pool of 100,000 to 1 billion (10^5-10^9) fixed compounds.

However, even a drop of water in the Pacific can't be counted in front of 60 out of 10 of these 1 billion molecules.

Even if it took countless manpower to find a few hit compounds, it would take years of long and painful iterative optimization to balance the molecule's toxicity, solubility and permeability.

That's why a new drug often takes 10 years, costs billions of dollars, and fails at more than 90 percent.

However, under the magic of Isomorphic Labs, "years" are greatly reduced to "days".

Max Jaderberg proudly displays their secret weaponryIsoDDE ((Isomorphic Labs engine for drug design)。

In the face of the same and even more difficult obstacles, IsoDDE has completely abandoned traditional methods of cooking.

It doesn't need to blindly filter out those poor billion inventories in reality, but to do a "smart search" for this endless 10^60 chemical space directly in the superbrain of a computer.

Just needA few daysIsoDDE is capable of autonomously generating and cyclically evolving quantities of molecules, designed to combine not only focal points, but also "perfect molecules" that are ready for real-world testing.

In traditional research and development, the long dead valley is seen as a thorn in the side of the sky and is directly flattened by AI.

In order to make the public intuitively aware of the terrifying power of AI, Jaderberg specifically outlined an optimised contrast to "Target D" (target D).

In this case, the only thing that a drug designer has to do is input a "demand manual" into the AI, such as which thresholds to hit, what regulation patterns to choose, how high selectivity, solubility and cell permeability standards to meet.

The rest will be handed over to the AI Design Agent.

In just two to four days, the AI intelligent body, like a tireless super-explorer, begins to explore wildly in chemical space.

If we look at this process on a visualized hotspot, we will see a very shocking scene:

Conventional molecular libraries screen for only a tiny "group" of areas on a heat map; while ISODDE's design intelligence is like laying out a huge network of radars, directly covering the most potentially desirable high-demand areas.

What's even more shocking is that AI understands.Dynamic Balance。

In the early exploration phase, its attention was very scattered, extensively testing various unprecedented chemical skeletons; once locked into high-potential areas, it immediately turned its attention and began deep targeted development (Exploit).

In the process, AI accurately controls the medication-designed holy grail."Pareto Frontier is the best".。

Anyone who makes drugs knows that the properties of molecules are often mutually exclusive.

For example, the stronger the drug, the more toxic it is; the better the solubility, the worse the permeability.

Traditional pharmacists painstakingly adjusted the tablet for years, while AI pushed directly above the multi-purpose Pareto boundary in a matter of days to find the "extreme molecules" that perfectly balance efficacy and bioavailability.

When researchers compared the AI-generated molecules with the reference molecules that human experts took 3 to 5 years to extract from the literature, the results were overwhelming. The functional concentration response curve demonstrates that the AI-designed molecules perform extremely well in all dimensions and are the perfect starting point for further development.

In this fifth anniversary speech, Jaderberg proposed a highly disruptive nuclear psychology:We are moving from the era of "experience medicine" to the era of "exquisitely engineered molecules".

Over the past hundred years, human medicine has remained essentially untouched by the experimental model of "God's own herb-tasting". Whether it be an accidental discovery of frosting or a targeted drug discovered through mass-produced blinding, it contains a large number of "lucky" components.

We just "coincidentally" encounter a molecule that cures a disease in nature or in an existing chemistry and try to fix it.

But in Jaderberg's eyes, the future of medicine shouldn't depend on luck, it should be like making planes, making chips."System engineering"。

In the dictionary, Isomorphic defines an "Exquisitely Engineered Molecule" as a highly volatile one: a compound from scratch that uses advanced AI designed with extreme precision to regulate biological focusing points according to established specifications and minimize toxicity.

In order to achieve this reduction-impact engineering capability, Isomorphic Labs has set up a high-levelVertical integrationIt's a big business.

Under this roof, the world's top AI researchers, biologists, chemists and clinicians are sitting together -- not only developing models, generating data themselves, discovering new biology, but delivering these potential drugs to the clinic themselves.

In 2024, AlphaFold 2 was awarded the Nobel Prize in Chemistry for solving the century-old problem of protein structure prediction.

Isomorphic Labs, as the AlphaFold bracket force, has been amplifying this revolutionary achievement indefinitely over the past five years.

The ISODDE they created is not just a structural prediction tool, but an end-engine that integrates "World Models" and "Frontier Reasoning Models".

The engine has a deep understanding of the physical and biological environment and a powerful zero-sample reasoning capability to simultaneously solve complex structural, bonding, molecular properties and toxicity problems.

Using AI to generate data, train models, lay out infrastructure, and ultimately mass-produce new drugs, Isomorphic is transforming drugs into a line of precision flow that can be calculated, predicted, and replicated.

What's even more shocking is that AI not only revolves around existing knowledge frameworks, it shows amazing "creativity".

Isomorphic Labs officially revealed that through IsoDDE, they have:

1. The new biological mechanisms have been discovered.
2. Designed a completely new molecule that opens "previously completely unknown protein pockets";
3. In real wet experiments, the unique biological functions of these molecules have been successfully verified.

This means that AI has gone beyond the stage of "learning humans how to make drugs" and is starting to reverse-guide humans: "Look, this hidden corner of this protein can be made into drugs, and I've even made the key molecules for you".

These molecules, rigorously verified by real biological experiments and demonstrated excellent pre-clinical data, are pushing Isomorphic Labs with an all-armed gesture to prepare for the final battlefield.Clinical trials and clinical developmentsStart the storm!

Looking back at the history of science, whether it be aerospace, automotive manufacturing or architectural design, almost every field of engineering has been completely disrupted by computer-aided design (CAD) in the last 50 years.

And today, the fire of disruption is finally burning in the most complex and sacred areas of humanity:Life and health.

This is the answer given by Isomorphic Labs, and it belongs to all human beings.

When the massive 60-dimensional chemical universe was scaled up by AI in a matter of days, it seemed to become tangible for the first time when drug development went from the "magic" of touching luck to the precise, repeatable "engineering" of a future without cure.

The "iPhone era" of AI pharmaceuticals has arrived. Five years is just the beginning, and the future of the pharmaceutical revolution will be even crazier and bigger than we ever imagined.