Isomorphic Labs, a spin-off of Google DeepMind, has introduced IsoDDE—a new drug design system capable of modeling the dynamic interaction between proteins and ligands.
What Happened
Developed by Isomorphic Labs, the IsoDDE system focuses on dynamic "induced fits" and cryptic binding sites, in contrast to the static approach of AlphaFold 3. In specialized tests, such as the Runs N' Poses benchmark, IsoDDE achieved a score of 50% compared to 23% for AlphaFold 3, and also demonstrated success in modeling complex antibody-antigen interactions.
Context
Traditional drug design methods face the challenge of understanding how molecules change shape upon binding. While AlphaFold 3 excels at predicting the static structure of a protein, IsoDDE moves toward modeling dynamics, which is critical for accurate molecular design in the early stages.
Why It Matters for the Industry
For the industry, this represents a technological leap in computational modeling that could significantly accelerate Phase I clinical trials (safety testing), where the success rate of AI-developed drugs already reaches 80-90%. However, the industry still faces the "Phase II wall" (efficacy testing), where the effectiveness of AI molecules is currently comparable to traditional methods, at approximately 40%.
Why It Matters for Users
For readers and specialists, this means AI has become much more efficient at predicting exactly how a drug will "stick" to a protein, accelerating the primary screening of candidates. At the same time, it is important to understand that the technology cannot yet guarantee an actual cure for a disease in a living organism without conducting full and expensive clinical trials.
What Is Still Unknown / Limitations
The main limitation remains the inability to predict the biological efficacy of a drug during Phase II clinical trials, where therapeutic action must be confirmed in living systems.
Sources
Author
Look at AI, Editorial Team