AlphaGenome is a newly released AI model from Google DeepMind that predicts the impact of genetic variants on gene regulation throughout the genome. It integrates long-range sequence context with base-pair resolution and can process up to 1 million base-pairs of DNA. This model enhances understanding of both common and rare variants in non-coding regulatory regions. AlphaGenome uses convolutional neural networks and transformers for improved performance in variant effect prediction, achieving superior results in 24 out of 26 evaluations. Additionally, it has enhanced training efficiency, requiring significantly less computational resources compared to previous models.
AlphaGenome processes up to 1 million base-pairs of DNA at once and outputs high-resolution predictions across thousands of molecular modalities, including gene expression and chromatin accessibility.
The architecture achieves state-of-the-art performance across a broad range of genomic benchmarks, outperforming task-specific models in 24 out of 26 evaluations of variant effect prediction.
Collection
[
|
...
]