Sciencepulse

Google's New AI Takes On the Genome's Dark Matter

Science·November 18, 2026

Google DeepMind has introduced AlphaGenome, an artificial intelligence model built to tackle one of genetics' longest-running blind spots: the vast stretches of DNA that never get translated into proteins. Researchers often call this noncoding material the genome's "dark matter." It makes up the large majority of our DNA, yet scientists still have a limited picture of how these regions influence gene expression, the process that determines when and where a gene is switched on. A tool that can read those signals more reliably could help labs work out how small changes in the genome translate into differences in health and disease.

On the medical front, doctors reported a striking case in which they kept a man alive for two days without working lungs, relying on artificial lungs to take over the job of oxygenating his blood. Support of this kind is usually a bridge, giving the patient's body time to recover or giving clinicians time to decide on the next step, so keeping someone stable for that long is a notable achievement in critical care.

The episode also looks at evolution. Researchers have cracked an evolutionary puzzle centered on a peculiar flower, nicknamed the rebel lipstick vine. Stories like this show how unusual forms in nature can carry clues about the pressures that shaped a species over long periods of time.

Taken together, the three stories share a theme: a better reading of the code, whether it is the genetic code, the body's limits or the history stored in a plant's form. AlphaGenome is still a research tool, and its predictions will need to be tested in real laboratories before they change how medicine is practiced. For now, it is a sign of how AI is being pointed at the parts of biology that scientists have struggled to interpret.