Curing Cancer in the Cloud: AI in Drug Discovery

The Eroom's Law

Moore's Law says computers get cheaper and faster every year. Eroom's Law (Moore's spelled backwards) says drug discovery gets slower and more expensive every year. The low-hanging fruit has been picked. Finding a new molecule that kills a disease without killing the patient is incredibly hard. The failure rate is 90%.

AI is reversing Eroom's Law.

The Protein Folding Problem

Proteins are the machines of life. Their function is determined by their 3D shape. If you know the shape (e.g., the spike protein of a virus), you can design a drug to lock into it (like a key in a lock). For 50 years, figuring out that shape took months of lab work (X-ray crystallography). AlphaFold (by DeepMind) solved this. It predicted the shape of 200 million proteins—effectively all known proteins/science. This gave biologists a map of the entire building blocks of life.

Generative Chemistry

Instead of screening a library of existing molecules, AI generates new ones. It's like ChatGPT, but for chemistry.

• Prompt: "Design a small molecule that binds to Receptor X, crosses the blood-brain barrier, and is not toxic to the liver."
• Output: A list of candidate molecular structures.

In 2023, the first AI-designed drug entered Phase 2 clinical trials for Pulmonary Fibrosis. It took 18 months to discover, compared to the industry average of 5 years.

Simulation over Experimentation

Before we give a drug to a mouse or a human, we test it on a Digital Twin. We simulate how the molecule interacts with the liver, heart, and kidneys. We kill the bad ideas in the cloud, so we don't harm patients in the real world.

Conclusion

Biology is becoming an information science. We are starting to understand the code of life well enough to debug it. AI is the tool that will allow us to cure the "incurable."