Google DeepMind 发布 AlphaGenome 2，预测DNA变异对特定细胞的影响
DeepMind 新版 AlphaGenome 2 在 28 种免疫细胞测试里把因果变异识别率从 16% 提到 30%，还能复现肠道炎症相关的已知发现。做药物靶点研究或计算生物学的人该看看。
Google DeepMind 发布 AlphaGenome 2，用于预测 DNA 变异如何影响特定细胞，帮助研究人员为新药寻找更好的靶点。同一处基因变异在不同细胞中可能产生不同效果，明确其作用位置有助于判断该研究哪个基因或生物学过程。
基准成绩上，在覆盖 28 种免疫细胞类型的测试中，AlphaGenome 2 识别出 30% 能因果性影响基因活性的变异，相比第一代模型改版后的 16% 有明显提升。它还复现了一条已知结论：某个 DNA 变异与肠道免疫细胞的炎症相关，为实验室验证潜在疗法提供了更具体的起点。
Beside all the frontier releases, this sadly gets too little attention. But its super significant what Google released today:
Google DeepMinds AlphaGenome 2 could help researchers find better targets for new drugs by predicting how DNA changes affect specific cells.
The same genetic change can have different effects depending on the cell. Knowing where it acts could help scientists identify which gene or biological process to investigate.
In a test across 28 immune cell types, DeepMind says AlphaGenome 2 identified 30% of variants that causally affect gene activity, compared with 16% for an adapted version of its first model.
It also reproduced a known finding linking a DNA variant to inflammation in gut immune cells. That gives researchers a more specific starting point for testing potential treatments in the lab.
So tl;dr AlphaGenome 2 could help researchers find better drug targets by predicting how DNA changes affect specific cells. Which is super helpful. And since many people asked me recently when we see real value created with the help of AI: Here you go guys.
Congrats @GoogleDeepMind !
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include AlphaGenome 2, Google DeepMind, 药物靶点, 生物信息.
Source attribution: curated from kimmonismus (@kimmonismus).
