DeepMind has unveiled AlphaGenome, a groundbreaking artificial intelligence tool poised to revolutionize the prediction of regulatory variant effects within DNA sequences. This innovative model can process exceptionally long DNA sequences, up to one million letters, providing high-resolution predictions for thousands of molecular properties that characterize gene regulatory activity. AlphaGenome distinguishes itself by offering comprehensive multimodal predictions, efficient variant scoring, and novel splice-junction modeling, effectively interpreting the vast non-coding regions of the genome that are crucial for gene orchestration. It achieves state-of-the-art performance across numerous genomic benchmarks, outperforming many specialized models in predicting DNA sequence and variant effects. DeepMind highlights AlphaGenome’s potential as a powerful research tool for understanding disease mechanisms, guiding synthetic biology, and accelerating fundamental genomic research, enabling scientists to explore a variant’s impact across multiple modalities with a single API call. Currently, AlphaGenome is accessible via an API for non-commercial research, with a broader release planned for the future, promising to deepen our understanding of the complex cellular processes encoded in DNA and drive new discoveries in genomics and healthcare.
https://deepmind.google/discover/blog/alphagenome-ai-for-better-understanding-the-genome/>