Google DeepMind发布AlphaGenome
Google DeepMind Releases AlphaGenome 2 for Cell-Type-Specific Variant Effects
做生物计算的同学注意，AlphaGenome 2解决了单细胞特异性预测难题，无需微调即可适配新细胞类型，这对理解基因调控机制很有价值。
Google DeepMind has released AlphaGenome 2, a DNA model that predicts variant effects in specific cell types. AlphaGenome 2 is a conditional sequence-to-function model. It reads 1 Mb of DNA plus a target cell state. Its predecessor averaged across bulk tissues. The new model resolves closely related cell types and adapts to unseen single-cell datasets without fine-tuning. It is available through the AlphaGenome API for non-commercial research.
Google DeepMind 发布了 AlphaGenome 2，这是一种 DNA 模型，可预测特定细胞类型中的变异效应。AlphaGenome 2 是一种条件序列到功能（sequence-to-function）模型。它读取 1 Mb 的 DNA 以及目标细胞状态。其前身是对混合组织（bulk tissues）进行平均处理。新模型能够解析密切相关的细胞类型，并适应未见过的单细胞数据集，而无需微调。该模型可通过 AlphaGenome API 用于非商业研究。
TL;DR
TL;DR（太长不看版）
- Size: Input is a 1 Mb (1,048,576 bp) DNA window, human or mouse.
- Runs on: Hosted API only for now. Training used 2048 TPU v4 chips; weights come “upon final publication”.
- Performance: Beats fine-tuned AlphaGenome 1, Borzoi and Scooby on single-cell eQTL and enhancer-gene benchmarks.
- Best: 43% recall at 90% sign accuracy on TenK10K sc-eQTLs, versus 23% for AlphaGenome 1 fine-tuned.
- Worst: Outside accessible chromatin, sign prediction AUROC reaches only 0.63.
- Bottom line: Best: cell-state conditioning from raw scRNA-seq. Worst: no local weights, non-commercial only.
- 规模：输入为 1 Mb（1,048,576 bp）的 DNA 窗口，适用于人类或小鼠。
- 运行环境：目前仅限托管 API。训练使用了 2048 个 TPU v4 芯片；权重将在“最终发表后”提供。
- 性能：在单细胞 eQTL 和增强子-基因基准测试中，优于经过微调的 AlphaGenome 1、Borzoi 和 Scooby。
- 最佳表现：在 TenK10K sc-eQTLs 上，当符号准确率达到 90% 时，召回率为 43%，而微调后的 AlphaGenome 1 仅为 23%。
- 最差表现：在可及染色质区域之外，符号预测的 AUROC 仅达到 0.63。
- 结论：最佳点：基于原始 scRNA-seq 的细胞状态条件调节。最差点：无本地权重，仅限非商业用途。
What is AlphaGenome 2?
什么是 AlphaGenome 2？
AlphaGenome 2 is a U-Net style model with a convolutional encoder, an 11-block transformer and a decoder. AlphaGenome 1 predicted a fixed catalog of thousands of tracks in parallel. AlphaGenome 2 instead predicts outputs for 1 condition per call.
AlphaGenome 2 是一个 U-Net 风格的模型，包含卷积编码器、11 层 Transformer 块和解码器。AlphaGenome 1 并行预测数千个固定目录轨道的输出。而 AlphaGenome 2 每次调用仅针对一个条件预测输出。
The condition is a 256-dimensional embedding. It modulates activations through FiLM layers in the encoder, transformer and decoder. Conditioning only the decoder or output layers gave minimal gains. Conditioning the encoder and transformer drove most of the improvement.
该条件是一个 256 维的嵌入向量。它通过编码器、Transformer 和解码器中的 FiLM 层调节激活值。仅对解码器或输出层进行条件调节带来的增益微乎其微。而对编码器和 Transformer 进行条件调节则带来了大部分改进。
Outputs cover RNA coverage, CAGE, splicing, chromatin accessibility, histone marks, TF binding and contact maps. CpG methylation is new.
输出涵盖 RNA 覆盖度、CAGE、剪接、染色质可及性、组蛋白修饰、TF 结合和接触图。CpG 甲基化是新增内容。
How was AlphaGenome 2 trained?
AlphaGenome 2 是如何训练的？
Training runs in 4 stages:
训练分为四个阶段：
- Pre-training across bulk data plus 3,977 single-cell RNA and ATAC pseudobulk tracks from 36 studies.
- Condition refinement, which tunes only the embeddings and FiLM layers.
- Distillation of 8 teacher models into 1 student, using mutated input sequences.
- Post-training that maps a single-cell expression encoder into the frozen model’s condition space.
- 在混合数据以及来自 36 项研究的 3,977 个单细胞 RNA 和 ATAC 伪批量轨道上进行预训练。
- 条件细化，仅调整嵌入向量和 FiLM 层。
- 将 8 个教师模型蒸馏为 1 个学生模型，使用突变输入序列。
- 后训练，将单细胞表达编码器映射到冻结模型的条件下空间。
That encoder was trained on over 100 million human scRNA-seq profiles, largely from CELLxGENE. Researchers can pass raw counts and get a condition embedding. The docs allow up to 256 cells per call.
该编码器是在超过 1 亿个人类 scRNA-seq 谱系上训练的，主要来自 CELLxGENE。研究人员可以传入原始计数并获得条件嵌入。文档允许每次调用最多处理 256 个细胞。
How does AlphaGenome 2 perform on benchmarks?
AlphaGenome 2 在基准测试中的表现如何？
On GTEx tissues, median Pearson r for cross-tissue expression changes rose from 0.61 to 0.71. Within tissues, explained cell-type variance rose from 23% to 34% for scRNA-seq.
在 GTEx 组织上，跨组织表达变化的中位数 Pearson r 从 0.61 上升至 0.71。在组织内部，scRNA-seq 解释的细胞类型方差从 23% 上升至 34%。
The main result is on TenK10K single-cell eQTLs across 28 immune cell types. At 90% precision, AlphaGenome 2 recovered 30% of causal variants. AlphaGenome 1 fine-tuned on single-cell data recovered 16%. Direction-of-effect recall rose from 23% to 43%.
主要结果来自 TenK10K 单细胞 eQTLs，涵盖 28 种免疫细胞类型。在 90% 精确度下，AlphaGenome 2 恢复了 30% 的因果变异。在单细胞数据上微调的 AlphaGenome 1 恢复了 16%。效应方向召回率从 23% 上升至 43%。
On held-out ENCODE-rE2G CRISPRi data, enhancer-gene AUPRC reached 0.78. AlphaGenome 1 scored 0.73, a Borzoi ensemble 0.61. The gap was widest for enhancers 100 to 2,500 kb away: 0.62 versus 0.48.
在未使用的 ENCODE-rE2G CRISPRi 数据上，增强子-基因 AUPRC 达到 0.78。AlphaGenome 1 得分为 0.73，Borzoi 集成模型为 0.61。差距在距离 100 至 2,500 kb 的增强子处最大：0.62 对比 0.48。
For assay imputation, AlphaGenome 2 beat the Avocado baseline on 166 of 181 held-out tracks. On an unseen Parkinson’s cohort, it reached 41% recall versus 17% for cohort-tuned Scooby.
对于实验数据插补，AlphaGenome 2 在 181 个未使用轨道中的 166 个上优于 Avocado 基线。在一个未见过的帕金森氏症队列中，其召回率达到 41%，而针对队列微调的 Scooby 仅为 17%。
How does it compare with other DNA sequence models?
它与其他 DNA 序列模型相比如何？
| Feature | AlphaGenome 2 | AlphaGenome 1 | Borzoi | Scooby |
|---|---|---|---|---|
| Input DNA | 1 Mb | 1 Mb | 524 kb | 524 kb |
| Condition handling | Reference or scRNA-seq embedding | Fixed multitask tracks | Fixed multitask tracks | Cell-embedding decoder, per dataset |
| Single-cell outputs | scRNA, scATAC | No | No | scRNA, scATAC |
| Parameters | Not disclosed | Not disclosed | 235M | Not disclosed |
| Hardware | Hosted API | H100 recommended | Not disclosed | NVIDIA GPU (tested on A40) |
| Weights / license | API now, weights later, non-commercial | Kaggle and HF, non-commercial terms | Apache-2.0 | MIT |
| TenK10K sc-eQTL causality AUPRC* | 0.76 | 0.72 (fine-tuned) | Not evaluated | 0.69 |
| TenK10K sc-eQTL sign AUC* | 0.85 | 0.79 (fine-tuned) | Not evaluated | 0.77 |
| rE2G held-out CRISPRi AUPRC* | 0.78 | 0.73 | 0.61 (4-model ensemble) | Not evaluated |
| 特征 | AlphaGenome 2 | AlphaGenome 1 | Borzoi | Scooby |
|---|---|---|---|---|
| 输入 DNA | 1 Mb | 1 Mb | 524 kb | 524 kb |
| 条件处理 | 参考或 scRNA-seq 嵌入 | 固定多任务轨道 | 固定多任务轨道 | 细胞嵌入解码器，每个数据集 |
| 单细胞输出 | scRNA, scATAC | 无 | 无 | scRNA, scATAC |
| 参数量 | 未披露 | 未披露 | 235M | 未披露 |
| 硬件 | 托管 API | 推荐 H100 | 未披露 | NVIDIA GPU（在 A40 上测试） |
| 权重/许可 | API 现用，权重后发，非商业 | Kaggle 和 HF，非商业条款 | Apache-2.0 | MIT |
| TenK10K sc-eQTL 因果性 AUPRC* | 0.76 | 0.72 (微调) | 未评估 | 0.69 |
| TenK10K sc-eQTL 符号 AUC* | 0.85 | 0.79 (微调) | 未评估 | 0.77 |
| rE2G 未使用 CRISPRi AUPRC* | 0.78 | 0.73 | 0.61 (4 模型集成) | 未评估 |
*Benchmark rows come from the AlphaGenome 2 paper and are DeepMind’s own evaluation.
*基准行数据来自 AlphaGenome 2 论文，是 DeepMind 自身的评估结果。
What can developers access today?
开发者今天可以访问什么？
AlphaGenome 2 and Colab tutorials are live through the API. An API key is free for non-commercial use. Commercial research is in private preview. The alphagenome Python package exposes the model under alphagenome.models.v2.
通过 API 已上线 AlphaGenome 2 和 Colab 教程。API 密钥可免费用于非商业用途。商业研究处于私有预览阶段。alphagenome Python 包在 alphagenome.models.v2 下暴露该模型。
DeepMind also released precomputed atlases. One covers enhancer-gene links across 640 primary cell conditions. Another scores about 7 million common and GWAS variants across 510 conditions. Both are on the downloads page.
DeepMind 还发布了预计算图谱。一个涵盖 640 种原代细胞条件下的增强子-基因连接。另一个对 510 种条件下约 700 万个常见变异和 GWAS 变异进行评分。两者均在下载页面提供。
Key Takeaways
关键要点
- AlphaGenome 2 conditions a 1 Mb DNA model on 1 cell type per call.
- Causal sc-eQTL recall at 90% precision rose from 16% to 30%.
- Raw scRNA-seq counts can define new cell states without retraining.
- Precomputed scores cover about 7 million variants across 510 conditions.
- Access is API-only and non-commercial for now.
- AlphaGenome 2 每次调用对 1 Mb DNA 模型施加 1 种细胞类型的条件。
- 在 90% 精确度下，因果 sc-eQTL 召回率从 16% 上升至 30%。
- 原始 scRNA-seq 计数可以在无需重新训练的情况下定义新的细胞状态。
- 预计算分数覆盖了510种条件下的约700万个变异。
- 目前仅支持API访问，且仅限非商业用途。
Check out the paper, the DeepMind blog, the API docs and the GitHub repo. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
请查阅论文、DeepMind博客、API文档和GitHub仓库。此外，欢迎在Twitter上关注我们，别忘了加入我们拥有15万+成员的ML SubReddit，并订阅我们的新闻通讯。等等！你在Telegram上吗？现在你也可以加入我们的Telegram群组了。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力