A quest for a cure: AI drug design with Isomorphic Labs
Experts from Isomorphic Labs discuss how AI, particularly models like AlphaFold3, is revolutionizing drug discovery. They explain the shift from the slow, iterative process of traditional methods to a new era where generative AI can design novel molecules, optimize for multiple biological properties simultaneously, and pave the way for personalized medicine.
The traditional process of drug discovery is incredibly complex, slow, and fraught with failure. It is an iterative game of “Whack-a-mole,” where solving one problem, like a molecule’s binding affinity, often creates another, like poor solubility. With a 90% failure rate from clinical trials to market, it’s a field accustomed to challenges. A significant hurdle is the reliance on animal models, which often fail to replicate human physiology accurately.
A New Era of Drug Design with AI
Artificial intelligence is set to fundamentally change this landscape. The process of drug design can be likened to “playing LEGO at the molecular level”: designing a small molecule (the drug) that fits perfectly into a crevice of a target protein to modulate its disease-causing function.
Models like AlphaFold3 have revolutionized this by predicting the 3D structure of these protein-molecule interactions in seconds—a task that previously took months or even years of lab work. This is accomplished using diffusion models that operate on a 2D interaction grid of all molecular elements (amino acids and atoms), generating precise 3D atomic coordinates for the entire system.
Generative Models and the Immense Chemical Space
The universe of potential drug-like molecules is astronomically large, estimated at 10^60 possibilities. Exhaustive searching is impossible. This is where generative AI becomes critical.
- Intelligent Exploration: Instead of brute-force searching, generative models and agents intelligently explore this vast chemical space to design novel molecules.
- Human-AI Collaboration: The models can suggest solutions that an expert chemist might not have conceived. In some cases, molecules designed by AI were indistinguishable from those designed by experienced medicinal chemists.
- Hypothesis Testing: This creates a powerful synergy. The AI provides novel hypotheses, and the human expert uses their judgment to decide which ones to test, accelerating the discovery process.
“In five-years’ time, doing drug design without AI will be like doing any sort of science without maths. And to be honest, I think the whole of science will be like, this. It’s like if you’re not using AI, what are you doing?” - Max Jaderberg
Beyond 3D Fit: A Multi-Objective Problem
A successful drug requires more than just a perfect fit. It must balance numerous, often conflicting, properties:
- Selectivity: The drug must bind to its target protein without affecting the other ~20,000 proteins in the human body, which could cause toxic side effects.
- Stability and Solubility: It must survive the acidic stomach environment and the metabolic processes of the liver, all while remaining soluble enough to be absorbed into the bloodstream.
AI models are being developed to predict and co-optimize for all these parameters from the very beginning of the design process. By building a “fingerprint of interactions” across the entire human proteome, these models can anticipate and design against potential toxicity and side effects in silico, long before human trials.
The Future: End-to-End Design and Personalized Medicine
The ultimate ambition is to create an end-to-end AI drug-design engine that can move from a specified disease target to a fully optimized drug candidate. This opens the door to personalized medicine, where drugs could be custom-designed for an individual’s specific genetic mutations, such as in a particular cancer.
While the vision of a single button that “solves all diseases” is still in the future, AI provides the tools to dramatically increase the speed and success rate of creating life-saving medicines.