← Latest briefing

Science

Google DeepMind launches AlphaGenome Atlas predicting impacts of human DNA variants

The petabyte-scale database models genetic mutations, though it remains far less accurate than AlphaFold.

The short version

  • Google DeepMind introduced AlphaGenome Atlas, a database charting predicted molecular effects across billions of human DNA variants.[The Verge · Scientific American]
  • Built on an AI model released last year that analyzes noncoding DNA, the petabyte-scale system aims to streamline laboratory workflows.[Scientific American]
  • DeepMind researchers hope the tool will spur medical research, though commercial entities such as drug developers must purchase a license.[The Verge · Scientific American]
  • While 30 times larger than the AlphaFold database, AlphaGenome Atlas is far less accurate in its current predictive performance.[Scientific American]

Key facts

  • Google DeepMind unveiled AlphaGenome Atlas, an AI-based database that charts predicted effects for nine billion single variants in the human genome.[The Verge · Scientific American]
  • The database relies on an AI model released last year focused on analyzing regulatory, noncoding DNA.[Scientific American]
  • The atlas encompasses a petabyte of data, making it 30 times larger than DeepMind's AlphaFold database.[Scientific American]
  • Commercial users, including pharmaceutical companies, are required to obtain a license to use AlphaGenome Atlas.[Scientific American]

What remains uncertain

  • AlphaGenome Atlas is acknowledged to be far less accurate than the AlphaFold database, and its real-world success in developing disease treatments remains to be proven.[The Verge · Scientific American]

Sources

Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.