DeepMind published AlphaGenome, a large sequence model designed to predict the functional effects of variants across the non‑coding 'dark' genome. The model takes long DNA contexts (up to one million base pairs) and outputs thousands of regulatory molecular properties, including transcription start sites, splicing and chromatin accessibility across tissues. AlphaGenome aims to prioritize causal non‑coding variants, speed target identification and guide synthetic regulatory design. The team says the model can score variant effects rapidly and is particularly geared toward rare variants and regulatory mechanisms that have been hard to interpret at scale. For genomics and drug‑discovery teams, AlphaGenome offers a new tool to triage regulatory variants from GWAS and to nominate mechanistic hypotheses for functional follow‑up, potentially shortening target validation timelines for diseases driven by regulatory dysfunction.