SEED
重启 HbF：松开 BACH2，也要保住 NRF2 Reopening HbF: relieve BACH2, preserve NRF2 人类遗传学揭示 BACH2–NRF2 调控轴：相邻启动子编辑可让 HbF 走向相反方向。
Human genetics reveals a BACH2–NRF2 axis in which neighbouring promoter edits can drive HbF in opposite directions.
Paper notes, opinionated essays, and a primer built from the ground up — keeping the parts of the field that actually hold up.
Hematopoietic stem cells, gene editing, and the road from bench to clinic.
How AI is starting to change the way biomedical research actually gets done.
人类遗传学揭示 BACH2–NRF2 调控轴：相邻启动子编辑可让 HbF 走向相反方向。
Human genetics reveals a BACH2–NRF2 axis in which neighbouring promoter edits can drive HbF in opposite directions.
研究者用配体感知的 de novo 蛋白设计，为三种 Janelia Fluor 染料造出小型、纳摩尔亲和力且彼此正交的 NovoTag，并把蛋白口袋进一步改造成寿命调节器、共价标签和染料诱导二聚系统。
Ligand-aware de novo protein design produced compact, nanomolar-affinity and mutually orthogonal NovoTags for three Janelia Fluor dyes, then turned their binding pockets into lifetime modulators, a covalent tag and a dye-induced dimerization system.
这篇尚未同行评议的预印本用 prime editing 给人 CD34+ 造血干细胞装上 CD117 E73K 护盾，再用 CIM058 清除未保护细胞，在人源化 β-地中海贫血小鼠中实现无 busulfan 的预处理与体内富集。
This unreviewed preprint uses prime editing to give human CD34+ cells a CD117 E73K shield, then pairs them with CIM058 to deplete unprotected cells and enrich the edited graft in a humanized β-thalassemia mouse model.
研究者把结构逆折叠与进化约束结合，设计出与天然 TnpB 显著分化、仍能在细菌、植物和人细胞中工作的 SynTnpB 基因编辑器。
Structure-conditioned inverse folding and evolutionary constraints generate sequence-divergent SynTnpB editors that remain active in bacterial, plant and human cells.
这项 Nature Genetics 研究用体内 Perturb-seq 并行扰动 79 个抑郁症 GWAS 风险基因，并在人类数据中找到一个共同指向神经元催产素 OXTR–ERK 信号下降的风险基因亚群。
An in vivo Perturb-seq screen of 79 MDD GWAS risk genes identifies a patient-matched subgroup converging on weakened neuronal oxytocin OXTR–ERK signaling.
Get one paper analysis in your inbox, every morning. No spam, unsubscribe any time.
CHIP 是一个筐,里面的克隆生长速度、被什么推着长、通向什么后果都不一样:DNMT3A 与 TP53 克隆平均每年约增长 5%,TET2、ASXL1、PPM1D 约 10%;已有研究把炎症、吸烟、化疗分别与不同的基因联系起来。对基因治疗而言,把所有 CHIP 克隆当作一个整体去追踪,会抹平这些差别。
CHIP is a basket: the clones inside it grow at different speeds, are pushed by different forces and lead to different outcomes. DNMT3A and TP53 clones grow about 5% a year on average, TET2, ASXL1 and PPM1D about 10%; existing studies link inflammation, smoking and chemotherapy to different genes. For gene therapy, tracking all CHIP clones as one group flattens these differences.
拆开 conditioning:一项 AML/MDS 随机试验里,减强度方案的复发率明显高于清髓方案;在镰刀型贫血成人的低强度移植研究里,多数患者仍稳定植入;小鼠实验则提示腾龛位是独立的一环。再看 busulfan 在基因治疗里怎样做剂量监测。
Pulling conditioning apart: in one randomized AML/MDS trial, relapse was clearly higher after a reduced-intensity regimen than after a myeloablative one; in a low-intensity transplant study in adults with sickle cell disease, most patients still engrafted stably; and mouse experiments suggest that making room in the niche is a separate step. Then, how busulfan exposure is monitored in gene therapy.
Arc、CZI、DeepMind 都在押注 AI 虚拟细胞。但把尺子对准最硬的那个结果：Arc 自己的挑战赛承认模型『还没能稳定超过朴素基线』，一篇 Genome Biology 评测里基础模型甚至输给『挑高变基因』。
Arc, CZI, and DeepMind are all betting on the AI virtual cell. But hold the ruler to the hardest result: Arc's own challenge admits models 'do not yet consistently beat naive baselines,' and a Genome Biology evaluation shows the foundation models lose to simply picking highly variable genes.
Doudna 团队用 AI 造出编辑效率反超天然酶的 SynTnpB，但论文标题里的 guided 才是最诚实的信号：AI 压缩的是酶设计这段可廉价验证的前端，通往体内安全那道贵门还没碰。
Doudna's lab built SynTnpB, an AI-designed nuclease that out-edits its natural template. But the honest signal is the word in the paper's title — guided: AI compressed the cheap front end of enzyme design, and never touched the expensive gate of in-vivo safety.
通用 agentic 科研平台会把“会用 AI 做科研”变成基础能力，真正的护城河会转向问题定义、证据判断和验证设计。
General agentic science platforms will turn AI-assisted research into baseline literacy; the moat shifts to question quality, evidence judgment, and validation design.