Isomorphic Labs: The $2.1B Bet to Solve All Disease #google #ai @IsomorphicLabs

本站添加: 2026-05-20

[00:00:00]Sir Demis Hassabis has raised $2.1 billion in Series B funding for Isomorphic Labs with a singular mission to use AI to solve all diseases. Right now, drug discovery is basically guesswork. We spend a decade and billions of dollars testing molecules, yet 95% of drug candidates fail in clinical trials. Isomorphic Labs is changing that by turning medicine into a software problem. Hassabis is the co-founder of Google DeepMind and a pioneer in AI who helped develop AlphaGo. You likely know him from AlphaFold, the Nobel Prize-winning AI that solved the 50-year-old protein folding problem. They open-sourced AlphaFold database and today over 3 million researchers use it to accelerate their hunt for new cures.

[00:00:36]Core bottleneck in pharma is efficiency.

[00:00:38]Moving from a biological target to a viable drug candidate traditionally takes 4 to 6 years. Most of these candidates eventually die in the clinic.

[00:00:45]The current industry workaround is high-throughput screening, which is essentially brute-forcing millions of chemicals in wet labs because traditional software does not generalize. It cannot reliably predict how a drug will behave in a novel biological system it has not encountered in its training data. Isomorphic's answer is IsoDE, their unified drug design engine. Instead of starting in a wet lab, IsoDE allows researchers to design medicines in silico, entirely on a computer. While previous models provided static photographs of proteins, IsoDE functions as an implicit world model. It understands the dynamic physics of induced fit, learning how proteins move, breathe, and twist when a drug interacts with them. On the rigorous running poses benchmark, IsoDE more than doubles the accuracy of AlphaFold 3 on the most challenging brand new systems. Most importantly, it can identify cryptic pockets, hidden binding sites that only appear when a protein reshapes itself around a molecule. In one major proof of concept, IsoDE computationally revealed a new allosteric pocket on the protein cereblon, a key target in cancer therapy. That experiment the scientists had missed for 15 years. This technology puts Isomorphic at the center of an AI drug discovery market projected to reach $4.5 billion by 2028. They've already secured nearly $3 billion in strategic partnerships with pharma giants like Eli Lilly and Novartis. These deals are deep multi-target collaborations aimed at the hardest problems in oncology and immunology. One is catch is that predictive fidelity is not the same as clinical cure. A molecule that looks perfect on the GPU must still survive the immense complexity of human toxicity and tumor heterogeneity. Additionally, regulators must decide how to evaluate drugs when the rationale for the design is embedded in the neural network rather than a human interpretable equation. To navigate this, Isomorphic utilizes a vertically integrated AI stack.

[00:02:28]AlphaFold provides a foundational structure. AlphaMissense classifies the pathogenicity of 89% of all human protein coding variants, and AlphaGenome illuminates the dark genome. The 98% of our DNA that controls how cells behave and where many cancer drivers are hidden. The result is a radically timeline compression. What once took years of trial and error can be front-loaded into months of AI-guided design.