How DeepMind's genome AI could help solve rare disease mysteries

"As part of an effort, called the Undiagnosed Hackathon, to crack 29 undiagnosed conditions researchers deployed AlphaGenome - an AI model developed by Google DeepMind in London that was described in Nature on 28 January. AlphaGenome - an AI tool that was made available to scientists last year - can predict the diverse effects of mutations in non-coding DNA sequences, including how they might affect the activity of nearby genes."

"Deciphering the 98% of the human genome that does not code for proteins is one of biology's grand challenges. Mutations in these sequences are especially vexing to researchers seeking to uncover the genetic basis for rare, often fatal diseases. "These are variants that, to be quite honest, often get triaged," says Eric Klee, a bioinformatician at the Mayo Clinic in Rochester, Minnesota, who co-led the Undiagnosed Hackathon in September last year."

More than 100 researchers convened at an Undiagnosed Hackathon to tackle 29 difficult medical conditions using artificial intelligence. They deployed AlphaGenome, an AI model from Google DeepMind that predicts effects of mutations in non-coding DNA and impacts on nearby gene activity. Deciphering the 98% of the genome that does not code for proteins remains a major challenge because mutations there complicate identification of genetic causes of rare, often fatal diseases. Around 350 million people have undiagnosed rare conditions, and genome sequencing currently diagnoses only a fraction. Diagnostic efforts often focus on protein-coding exomes, so non-coding variants are frequently triaged. The Wilhelm Foundation organized the hackathons to advance diagnoses for affected families.

Read at Nature