genomics

The paper introduces CUTTIE, a hierarchical testing method that clusters a trained network's units into a tree and tests them top-down with conditional independence tests to localize predictive concepts (e.g., regulatory DNA motifs) while controlling the family-wise error rate, demonstrating it on a model predicting chromatin accessibility from sequence.

HuggingFace Bio 团队发布 FineDNA——迄今最大的 DNA/基因组学 LLM 训练数据集，包含 300B+ token 的基因组与 DNA 相关文本、约 20 万条 SFT 对话和约 9.8 万条可验证 RL 提示，以及将 Qwen3-14B 训练成最强开源 DNA 模型 Carbon-X 的完整代码与配方。

A new preprint from Arc Institute reports that Evo 2 can design enhancers that label specific neuronal subtypes in live mouse brains, using fewer than 100 example sequences as prompts, advancing programmable gene expression for biology and therapeutic engineering.

A large China-based collaboration released Pinal, an AI tool that turns natural language prompts into functional protein designs, while Hugging Face launched a similar tool for DNA-level genome design—bringing 'vibe coding' style accessibility to protein and genome engineering.

HuggingFaceBio 发布开源 14.8B DNA 理解与生成模型 Carbon-X，基于 Qwen3-14B 微调，融合 DNA 序列与自然语言训练，在五项 DNA 基准上平均得分 66.7%，接近前沿闭源模型但推理成本约为其 1/1800。

Hugging Face released Carbon-X, an open DNA language model that answers questions about gene sequences in plain English, scoring 66.7% on DNA benchmarks near GPT-6.1 Sol at roughly 1/1,800 the cost per problem. It runs on a single 80GB GPU, potentially making large-scale variant classification dramatically cheaper for hospitals and researchers, though it is not yet clinically validated.

Carbon-X is an open 14.8B decoder-only model that jointly processes raw DNA and text via a shared 6-mer token stream, achieving 66.7% average on five DNA benchmarks — nearly matching GPT-6.1 Sol — with Apache 2.0 weights that fit on a single 80GB GPU.

HuggingFace Bio 发布 Carbon-X，一个 14B 的 DNA+自然语言联合模型，在五项 DNA 基准上超越所有开源模型并匹配 GPT-6.1 Sol，且成本低约 1000 倍，同时提供 GGUF 版本支持 llama.cpp 本地运行。

Hugging Face released Carbon-X, an open model that lets users chat with DNA in natural language, trained on 300 billion tokens of genomic sequences and English text to help researchers interpret genetic mutations and design therapies.

Arc Institute 的 Science Fellow 周景天及其团队在新预印本中介绍 Bolero，这是一个以细胞状态为条件、仅从 DNA 预测染色质可及性与基因表达的模型，解释了相同 DNA 在不同细胞中的差异解读。

文章比较了 DNA 语言模型（如 Evo 2）与 AlphaGenome 在预测 DNA 变异影响上的差异：前者基于序列似然评分，后者基于实验测量预测基因表达与剪接等读出，且仅限研究用途。

Wired/Quanta 报道指出，AI 基因组基础模型（如 Evo 2、Genos、AlphaGenome）仍无法解读人类基因组复杂的三维物理结构及其基因调控机制，说明 AI 目前尚不足以真正破解我们的基因组。

Google DeepMind highlights how its AI models and tools—including AlphaFold and AlphaGenome Atlas—are helping scientists worldwide make diseases more detectable, treatable, and preventable, from drug repurposing to mapping genetic variants linked to aging.

论文提出 AutoOnboarder，一个由智能"Onboarder"编码代理监督的自主系统，通过迭代调整基础模型、提示词和 harness 源代码来适配任务，在基因组测序预授权表单提交任务中以最高低 29 倍的成本匹配或超越 Gemini 3.5 Flash 基线表现。

Pythia, an interpretable AI model trained on a new high-resolution atlas of 67 RNA-binding proteins, is introduced as a collaboration between Genophoria, CZ Biohub, UCSF, and Calico to predict how RBPs recognize RNA by sequence and shape.

Google 聚合其五年来在医疗健康领域的研究成果，涵盖 AlphaGenome Atlas、AlphaFold 药物筛选、脑科学纳米级重建等，展示了 AI 如何加速疾病检测、治疗与预防。

AskCyph Health Explorer 是一款教育工具，结合 Google DeepMind 的 AlphaGenome API 预测基因变异影响，以及在浏览器中通过 WebGPU 本地运行的 MedGemma 1.5 4B 模型，实现完全隐私的 DNA 突变与疾病关系探索。

Two independent research teams solved the 60-year mystery of orange cat fur, identifying a 5kb DNA deletion near the Arhgap36 gene on the X chromosome that drives the trait and explains its sex-linked inheritance patterns.

DeepMind has released AlphaGenome, a machine learning model that predicts the impact of genomic variants at base-pair resolution with a context window of one million base pairs. It matches or outperforms the strongest competing models on 25 of 26 variant effect benchmarks. Accompanied by the AlphaGenome Atlas, which has pre-computed the effects of roughly nine billion single-nucleotide variants — a scale 30 times that of the AlphaFold database — the model weights are permanently free for the scientific community.

科学家首次对世界上现存最古老的陆地动物——约194岁的巨龟Jonathan进行基因组测序，发现其在DNA修复、胰岛素调节、线粒体功能等衰老通路中拥有独特基因变异，相关研究发表于《Science Advances》。