AlphaGenome AI Model Expands Genomic Research Capabilities

AlphaGenome is a deep-learning system developed by Google DeepMind to support research into how DNA segments control gene expression and how genetic variants may affect that regulation. The tool is intended for research purposes and is not validated for clinical use.

The model was released on 25 June 2025 and can be accessed through an API for non-commercial research projects. AlphaGenome was trained using several large, publicly available multi-omic datasets, including ENCODE, GTEx, 4D Nucleome, and FANTOM5.

AlphaGenome accepts DNA sequences up to one million base pairs in length. It generates thousands of quantitative predictions related to various regulatory modalities, such as gene expression, chromatin accessibility, splicing, transcription factor binding, and chromatin contacts.

In benchmarking tests, AlphaGenome matched or surpassed other leading models in a majority of evaluated tasks. The model demonstrated strong performance in both genomic track prediction and variant effect prediction benchmarks.

What the numbers show

• Launched on 25 June 2025
• Processes DNA sequences up to one million base pairs
• Outperformed or matched state-of-the-art models in 22 of 24 genomic track prediction tasks
• Matched or exceeded benchmarks in 24 of 26 variant effect prediction tests

AlphaGenome was able to replicate a known regulatory mechanism in T-cell acute lymphoblastic leukemia during testing. The model predicted how a specific non-coding mutation could activate the TAL1 oncogene, providing an example of its application in research settings.

Despite its capabilities, AlphaGenome has some limitations. It does not capture regulatory effects from DNA variants located more than 100,000 base pairs away and does not offer personalized or clinical prediction features.

The tool is designed for non-commercial research and is not intended for use in clinical decision-making. Its predictions are based on patterns learned from large-scale public datasets rather than individual patient data.

AlphaGenome represents an example of how artificial intelligence can be applied to genomic research, offering researchers a resource for analyzing the regulatory impact of genetic variants across large DNA regions.

* This article is based on publicly available information at the time of writing.