Recently, when Google Deepmind CEO Demis Hassabis won the Nobel Prize, he didn't win it because of physics. Instead, he won over chemistry, especially the type of chemistry that enters our body within our genes and DNA. Currently, DeepMind is introducing a new artificial intelligence model, alphagenome. This is a novel AI model specifically tailored to accurately predict how individual mutations in human DNA affect function.
In other words, AI can help scientists and doctors better understand genomic function. This advanced AI model, developed by DeepMind, is said to bring about a major leap in genome research. The human genome is a complete set of genetic instructions. Think of it as a comprehensive instruction manual with data to build and operate living organisms. The genetic material is mainly made up of DNA. Genomes can affect everything from human physical properties to potential risks of diseases like cancer.
Small changes in our DNA can have a major impact on our health. But now, understanding genes and how these changes work at a molecular level is one of the biggest challenges for biologists. DeepMinds Alphagenome AI aims to help researchers solve these answers by providing deeper insights into genomic mechanisms, particularly in parts that do not directly encode proteins and play an important role in regulating genes.
How the alphagenome works
Deepmind is the ability to assess both general and rare genetic variants, which are small changes in DNA that make each person unique, that is, in its core, alphagenomes are particularly unique. According to the company, this was made possible by major technological advances in which the model analyzes very long DNA sequences (up to 1 million base pairs) and generates highly detailed predictions. More importantly, AI models can do this within all models with many different cell types and biological processes.
DeepMind reveals that the new alphagenomes were trained using extensive public datasets from large consortiums such as encoding, GTEX, 4D Nucleome, and Fantom5, which closely measured these properties in a large number of human and mouse cell types.
Up until now, researchers have often relied on multiple tools to study how gene mutations affect various aspects of gene regulation. However, according to DeepMind, alphagenomes change this process. Several features are combined into one model to reduce the need for fragmented approaches, allowing for faster and more comprehensive analysis.
In particular, Alphagenome is built on the early genomics model of Google Deepmind, Enformer, and complements Alphamissense, which specializes in analyzing variants within 2% of protein-encoding genomes. The company emphasizes that alphagenomes are essential for regulating gene activity and provide an important new perspective to interpret the vast 98% (non-coding region), including many variants associated with the disease.
The company emphasizes that researchers are already using alphagenomes to investigate the possibility that certain genetic mutations can lead to cancer. One test accurately predicted how leukemia-related mutations activate harmful genes, confirming previous experimental findings.
Availability of alphagenomes
Alphagenomes are currently available through APIs for non-commercial and research-centric use. Although no clinical diagnosis has been approved, Google says AI tools can help scientists identify which mutations are most likely to cause the disease. However, the company notes that the model is still evolving and future versions may cover more species, cell types, or biological processes.
– end
