Google DeepMind published AlphaGenome, a sequence‑based model that predicts regulatory effects of non‑coding genomic variants across long DNA contexts up to one million base pairs. The model outputs thousands of molecular properties — such as DNA accessibility, gene start sites and splicing predictions — enabling rapid scoring of variant consequences for functional genomics and target discovery. AlphaGenome’s creators position the model as a tool for prioritizing rare regulatory variants and interpreting genome‑wide association signals, with potential applications in therapeutic target identification and synthetic DNA design. Performance and speed claims include single‑second variant scoring across multiple modalities, which could accelerate variant interpretation in research and clinical contexts. Nature published the work and independent experts noted AlphaGenome’s potential to help triage non‑coding variants, a longstanding bottleneck in human genetics and drug target validation.