Is AI the Future of Drug Discovery?

Deep Genomics, an AI drug discovery play, announced that it generated a drug candidate for Wilson’s Disease, a rare metabolic condition that impairs copper metabolism.

This is a promising advance in some respects. It may very well result in a treatment for Wilson’s sufferers. It validates the power of AI in drug discovery. But it also shows why AI won’t make an impact on human health.

Structure of ATP7B protein, a copper-transporting P-type ATPase which is encoded by the ATP7B gene. By ProteinBoxBot at English Wikipedia. [Public domain]

Here are the particulars:

Wilson’s disease is an inability to regulate copper levels in the body. As with so many things, a small amount of copper is essential to good health, but too much is toxic. The disease is known to be due to variations in a specific gene (ATP7B), but scientists were unable explain how these mutations gave rise to the disease.

Deep Genomics’ AI analysis may have solved this mystery. It suggests that one particular mutation causes disease by errors in the production of the protein, rather than by aberrant activity of the protein itself. DG did some experiments on cells grown in isolation that support this hypothesis.

DG then used AI to generate drug candidates. But the therapeutic platform they searched was not what you might think of as a drug. They didn’t come up with a small molecule that can be taken as a pill, like a statin, nor did they identify a biological molecule that…