Google DeepMind has introduced AlphaGenome, an AI tool that advances our understanding of how genetic variants impact biological processes. This model processes up to 1 million DNA letters and predicts thousands of molecular properties regulating genes. Published in Nature, AlphaGenome represents a significant leap in genomic research by offering more comprehensive predictions than previous models.

AlphaGenome analyzes DNA sequences with high resolution, capturing fine-grained biological details that previous models missed. The tool predicts where genes start and end in different cell types, splicing patterns, RNA production levels, and protein binding sites. By comparing mutated and unmutated sequences, it efficiently scores variant effects across multiple molecular properties in just one second.

State-of-the-art performance benchmarks show AlphaGenome outperformed existing models on 22 of 24 evaluations for single DNA sequences and matched or exceeded top models on 24 of 26 variant effect predictions. Its generality allows scientists to explore a variant's impact on multiple modalities simultaneously through a single API call, accelerating hypothesis generation and testing in genomic research.