Introduction of AlphaGenome 2 by Google DeepMind
Google DeepMind has recently unveiled AlphaGenome 2, an innovative artificial intelligence model poised to transform scientific research by serving as “a new and powerful tool for discovering the cellular factors that drive diseases,” as articulated by its creators. This cutting-edge DNA sequencing model integrates single-cell biology with genetics, aiming to predict how genetic variants influence the behaviour of hundreds of types of human cells.
Understanding Genetic Expression in Human Cells
Scientific literature indicates that nearly all cells within the human body share the same genetic code. However, the expression of genes varies significantly among different cell types. Each cell activates or deactivates specific groups of genes based on its particular function. This process, known as gene expression, is crucial in determining which genes are active and the roles each cell plays within the body.
Previous initiatives, such as the AlphaGenome Atlas, have facilitated the identification of the behaviour of millions of cells. Nevertheless, a complete understanding of how minor DNA modifications affect cellular activity and the implications of these changes for the emergence of certain diseases or the development of particular human traits remains elusive. AlphaGenome 2 aims to bridge this knowledge gap by identifying how small variations in genetic sequences can alter gene activity in specific cells, thereby paving the way for a deeper understanding of the biological mechanisms associated with various diseases.
How AlphaGenome 2 Operates
The functionality of AlphaGenome 2 builds upon the advancements made by its predecessors, AlphaGenome and AlphaGenome Atlas. The former is designed to predict how genetic variants impact molecular activity across different tissues, while the latter compiles predictions regarding the 9 billion potential single-letter changes within the human genome. Together, these developments empower researchers to identify and investigate genetic variations that may contribute to rare diseases, complex human traits, and fundamental biological processes.
The Google team emphasises that “these tools assist researchers in prioritising and interpreting the effects of variants, thereby propelling discoveries in rare diseases, complex traits, and fundamental biology.” A significant enhancement of AlphaGenome 2 is its ability to analyse individual cell types and states rather than merely assessing the average behaviour of genes within entire tissues.
Advancements in Genetic Variant Identification
Constructed from the foundational elements of AlphaGenome, the new model features a redesigned architecture and a training process that utilised over 100 million profiles from individual cells across 68 human tissues. As a result of this update, the team asserts that AlphaGenome 2 nearly doubles the capacity of its predecessor in identifying genetic variants that modify gene expression in cells.
In practical terms, the model enables researchers to predict the effects of variants in disease-relevant cells that may be overlooked in general tissue models, as well as to generate predictions for cellular states and diseases that were not previously considered. This advancement permits the utilisation of cellular atlases not merely as descriptive maps of cell characteristics but also as predictive resources concerning human genetics.
Availability for Research Purposes
AlphaGenome 2 is available for non-commercial research through its API, while access for commercial research projects is currently in a preliminary private phase. This accessibility aims to foster collaboration and innovation in the scientific community, enhancing the potential for groundbreaking discoveries in the realm of genetics and disease research.