siRNA 设计指南siRNA Design Guide
从靶点选择、序列筛选、脱靶检查到 qPCR 与 Western 验证——第一次设计 siRNA 就能走完的完整流程。本书解决的问题：拿到一个基因名之后，不知道选哪段序列、怕脱靶、订了 3 条全都不敲低。Target selection, sequence screening, off-target checking, and validation by qPCR and Western blot — the complete workflow for your first siRNA design. The problem this guide solves: you have a gene name, but no idea which stretch to pick, a fear of off-targets, and three ordered siRNAs that knock down nothing.
适用对象：第一次独立设计 siRNA 的实验人员（细胞实验，体外转染）Audience: first-time siRNA designers (cell culture, in-vitro transfection)
前置要求：会查 RefSeq / Ensembl，会做基本转染和 qPCRPrerequisites: can use RefSeq / Ensembl; basic transfection and qPCR
版本 v1.0 · 2026 年 11 月Version 1.0 · November 2026
先记住三个数字：19–21 nt 是经典 siRNA 双链区长度（另加 3′ 端 2 nt 悬垂，常用 dTdT）；30%–52% GC 是候选序列的 GC 窗口；mRNA 水平敲低 70% 以上才算进入可用候选，最终至少保留 2 条独立序列互相印证表型。Three numbers to remember: 19–21 nt is the classic siRNA duplex length (plus a 2 nt 3′ overhang, usually dTdT); 30%–52% GC is the candidate GC window; and ≥ 70% knockdown at the mRNA level makes a candidate usable — keep at least 2 independent sequences that corroborate the phenotype.
最常见的失败原因不是序列，是流程：只订 1 条、没有阳性对照、只测 mRNA 不测蛋白、转染没优化就怪序列。本指南每一章都在堵这几个口。第 4 章的脱靶检查和最后的订购前清单是硬门槛，没过不要下单。The most common failure is workflow, not sequence: ordering only one siRNA, no positive control, measuring mRNA but not protein, blaming the sequence before optimizing transfection. Every chapter plugs one of these holes. The Chapter 4 off-target checks and the final pre-order checklist are hard gates — do not order until you pass them.
第 1 章 · RNAi 原理简介Chapter 1 · How RNAi Works
1进入细胞：转染 siRNA 双链Entry: siRNA duplex transfected
2装载 RISC：Ago2 挑中一条链RISC loading: Ago2 keeps one strand
3丢弃过客链并降解Passenger strand discarded
4guide 链搜寻靶 mRNAGuide strand finds target mRNA
5Ago2 在第 10–11 位间切断Ago2 cuts between positions 10–11
6mRNA 降解，蛋白随之下降mRNA degraded, protein falls
链选择不对称。RISC 不是随机留链：双链两端稳定性不同时，Ago2 倾向留下 5′ 端更不稳定（A/U 多）的那条作 guide——这是第 3 章"热力学不对称性"的来源。切割位点固定在 guide 链 5′ 端数起第 10–11 位之间，所以靶序列中间区域的错配最致命。seed 区（guide 第 2–8 位）只需部分配对就能压制一批其他 mRNA，是脱靶的主要来源（第 4 章）。Strand selection is asymmetric. RISC does not keep a random strand: when the two duplex ends differ in stability, Ago2 prefers the strand whose 5′ end is less stable (more A/U) as the guide — the origin of the thermodynamic-asymmetry rule in Chapter 3. The cut site is fixed between positions 10–11 counting from the guide's 5′ end, so mismatches in the middle of the target hurt most. The seed region (guide positions 2–8) needs only partial pairing to repress many other mRNAs — the main source of off-target effects (Chapter 4).
术语：正义链叫 passenger strand（过客链），反义链叫 guide strand（引导链）。序列除特别注明外按 5′→3′ 书写；靶序列指 mRNA 上与 guide 链互补的那段 19 nt。siRNA 是化学合成双链、直接转染，起效快（24–72 h）但短暂；shRNA 由质粒/慢病毒表达，持久，适合长期敲低；miRNA 模拟物多靶点、部分抑制，一般不切割。Terminology: the sense strand is the passenger strand; the antisense strand is the guide strand. Sequences are written 5′→3′ unless noted; the target sequence is the 19 nt stretch of mRNA complementary to the guide. siRNA is a synthetic duplex, transfected directly — fast (24–72 h) but transient. shRNA is expressed from plasmid/lentivirus — durable, for long-term knockdown. miRNA mimics hit many targets with partial repression and usually no cutting.
第 2 章 · 靶点选择：选哪个区域、避开什么Chapter 2 · Target Selection: Where to Aim, What to Avoid
序列优劣是第二位的，位置第一。先定转录本再定区域：在 RefSeq 或 Ensembl 确认要敲哪个转录本；敲全部亚型就选共有外显子，只敲单一亚型才选其独有外显子或剪接连接处。优先编码区 CDS，一般从起始密码子下游约 50–100 nt 后开始找；避开 5′ 端紧邻起始密码子的区域（翻译机器与 RNA 结合蛋白密集，效果常偏弱）。后续要做 rescue 实验时，选 3′UTR 或可做同义突变的 CDS 位点更方便。Position beats sequence. Fix the transcript first, then the region: confirm the target transcript in RefSeq or Ensembl. To knock down all isoforms, aim at exons shared by all of them; aim at a unique exon or splice junction only when you want one isoform. Prefer the CDS, starting roughly 50–100 nt downstream of the start codon; avoid the region immediately around the start codon, where translation machinery and RNA-binding proteins crowd in and efficacy is often weak. If you plan a rescue experiment, a 3′UTR site or a synonymously mutable CDS site is more convenient.
| 避开对象Avoid | 为什么Why |
|---|
| 已知 SNP 位点Known SNP sites | 你的细胞系若是另一等位基因，错配直接让敲低失效；查 dbSNPIf your cell line carries the other allele, the mismatch kills knockdown; check dbSNP |
| 可变剪接差异区（除非故意）Alternatively spliced regions (unless deliberate) | 靶在被跳过的外显子上等于没靶A target on a skipped exon is no target at all |
| 低复杂度 / 重复序列Low-complexity / repetitive sequence | 脱靶面大且合成质量差Broad off-target surface and poor synthesis quality |
| 同源基因高度相似区Regions shared with homologous genes | 家族成员会被一起敲，分不清表型归谁Family members get knocked down together; the phenotype cannot be attributed |
| 强二级结构区（有条件时）Strong secondary structure (when predictable) | mRNA 折叠紧的区域可及性差Tightly folded mRNA is poorly accessible |
分散布点：3–4 条候选不要挤在相邻位置，沿 CDS 前、中、后分散选点——某个区域敲不动时其他位置还有机会，多条不同位置 siRNA 给出同一表型也是排除脱靶最有力的证据。Spread your picks: do not cluster 3–4 candidates in adjacent positions; spread them along the CDS (front, middle, back). If one region resists knockdown, others still have a chance — and the same phenotype from siRNAs at different positions is the strongest evidence against off-target effects.
第 3 章 · 设计规则与评分表Chapter 3 · Design Rules and the Scoring Table
| 硬规则Hard rule | 标准与原因Standard and reason |
|---|
| GC 含量 30%–52%GC content 30%–52% | 过低结合不稳；过高双链解不开、装载差、脱靶增加Too low: weak target binding. Too high: duplex won't unwind, loading suffers, off-targets rise |
| 无长同碱基连排No long homopolymer runs | 避免 4 个以上相同碱基连续（AAAA、GGGG）Avoid 4+ identical bases in a row (AAAA, GGGG) |
| 无明显回文 / 强内部发夹No strong palindromes / internal hairpins | 序列自身折叠会与靶结合和装载竞争Self-folding competes with target binding and loading |
| 避开免疫刺激基序Avoid immunostimulatory motifs | 如富 GU 序列可激活 TLR，引起非特异反应GU-rich motifs can trigger TLRs and non-specific responses |
| 3′ 端加 2 nt 悬垂Add a 2 nt 3′ overhang | 经典用 dTdT，利于 RISC 识别；悬垂不计入 19 nt 匹配区Classically dTdT; aids RISC recognition; the overhang is outside the 19 nt match region |
热力学不对称性（本章最重要的一条）：目标是让 guide 链被选中——让双链在 guide 链 5′ 端一侧更不稳定：guide 链 5′ 端第 1 位偏好 A 或 U，5′ 端前 4–5 个碱基中 A/U 占多数；另一端相对富 G/C。一句话：guide 的头要"松"，尾要"稳"。Thermodynamic asymmetry (the most important rule in this chapter): the goal is to get the guide strand selected — make the duplex end at the guide's 5′ side less stable: position 1 of the guide prefers A or U, A/U should dominate its first 4–5 bases, and the other end should be relatively G/C-rich. In one line: the guide's head should be loose, its tail firm.
| # | 评分项（每满足一项记 1 分）Scoring item (1 point each) | 来源Source |
|---|
| 1 | GC 含量在 30%–52% 之间GC content within 30%–52% | Reynolds |
| 2 | 正义链第 15–19 位中至少 3 个 A/UAt least 3 A/U among sense-strand positions 15–19 | Reynolds |
| 3 | 正义链第 3 位为 ASense position 3 is A | Reynolds |
| 4 | 正义链第 10 位为 USense position 10 is U | Reynolds |
| 5 | 正义链第 13 位不是 GSense position 13 is not G | Reynolds |
| 6 | 正义链第 19 位不是 G/CSense position 19 is not G/C | Reynolds |
| 7 | guide 链 5′ 端为 A/U，该端不稳定（不对称性）Guide 5′ end is A/U and that duplex end is unstable (asymmetry) | Ui-Tei / Schwarz |
| 8 | guide 链 5′ 端 7 nt 内 A/U 占优A/U dominate the guide's first 7 nt | Ui-Tei |
| 9 | 不含 9 nt 以上的 GC 连续 stretchNo GC stretch longer than 9 nt | Ui-Tei |
Reynolds 原始研究中得分 6 分以上的候选成功率明显更高。但不要迷信单项规则——没有任何规则能保证单条必中，这正是每个项目订 3–4 条的原因。分数用来排序，不用来把候选砍到只剩一条。也不要直接抄文献或商品 siRNA 的序列而不核对转录本版本。In Reynolds' original study, candidates scoring 6+ succeeded markedly more often. But never trust any single rule — none can guarantee one siRNA will work, which is exactly why every project orders 3–4. Use the score to rank, not to cut down to one. And never copy a published or commercial siRNA sequence without re-checking the transcript version.
第 4 章 · 脱靶检查：BLAST 与 seed 评估Chapter 4 · Off-Target Checking: BLAST and Seed Analysis
脱靶分两种，检查方法完全不同，只做全长 BLAST 就宣布"无脱靶"是新手最常见的错误。近全长匹配脱靶：把 19 nt 靶序列在 NCBI BLAST 中对目标物种转录本库做 blastn 短序列模式（word size 调小如 7，关低复杂度过滤）；除目标外若有转录本 15 nt 以上连续匹配，或总错配 ≤2 且不在切割位点附近，候选淘汰或降级，特别注意基因家族成员和假基因。seed 区脱靶：取 guide 第 2–8 位 7 nt，查它在全转录组 3′UTR 的匹配频率；seed 匹配越多潜在被压制的无关基因越多，seed 区 GC 过高时风险额外升高。完全避开不现实，目标是选 seed 匹配谱相对干净、且几条候选 seed 互不相同的组合。Off-targets come in two kinds with completely different checks; running only a full-length BLAST and declaring "no off-targets" is the most common beginner mistake. Near-full-length matches: BLAST the 19 nt target against the species' transcript database in short-sequence mode (word size ~7, low-complexity filter off). If any other transcript matches ≥ 15 nt contiguously, or has ≤ 2 mismatches away from the cleavage site, drop or demote the candidate — watch gene-family members and pseudogenes especially. Seed-mediated off-targets: take guide positions 2–8 (7 nt) and check its match frequency across transcriptome 3′UTRs; more matches mean more unrelated genes potentially repressed, and a GC-rich seed raises the risk further. Zero seed off-targeting is unrealistic — aim for a relatively clean seed profile and candidates whose seeds all differ from each other.
实验层面兜底比预测更重要：2 条以上 seed 完全不同、位置不同的 siRNA 给出同一表型，再加上 rescue 实验，才是排除脱靶的标准证据链。任何预测工具都替代不了。Experimental backup beats prediction: the standard evidence chain against off-targets is two or more siRNAs with completely different seeds and positions giving the same phenotype, plus a rescue experiment. No prediction tool replaces it.
第 5 章 · 化学修饰简介：先问需不需要Chapter 5 · Chemical Modifications: Ask Whether You Need Them
| 修饰Modification | 作用Purpose | 注意Caution |
|---|
| 2′-O-methyl (2′-OMe) | 抗核酸酶、降低免疫激活；放在 seed 区特定位置可减轻 seed 脱靶Nuclease resistance, less immune activation; placed in the seed it can reduce seed off-targets | guide 链全长重修饰会损害装载与切割Heavy modification along the whole guide harms loading and cleavage |
| 2′-fluoro (2′-F) | 提高稳定性与亲和力，常与 2′-OMe 交替（体内药物常用）Stability and affinity; often alternated with 2′-OMe in in-vivo drugs | 体外实验性价比低Poor value for in-vitro work |
| 硫代磷酸酯（PS）Phosphorothioate (PS) | 骨架抗降解、促进蛋白结合与体内分布Backbone protection, protein binding, in-vivo distribution | 过多增加非特异结合与毒性Too much raises non-specific binding and toxicity |
| 5′ 端磷酸5′ phosphate | guide 链 5′ 磷酸是 Ago2 识别要素The guide 5′ phosphate is an Ago2 recognition element | 合成时向厂商确认 5′ 状态，别默认Confirm the 5′ status with the vendor; never assume |
| dTdT 悬垂overhang | 3′ 端 2 nt 脱氧悬垂，经典设计Classic 2 nt deoxy 3′ overhang | 也可做 UU 等 RNA 形式UU and other RNA forms also work |
按场景决策：体外转染短期实验用标准未修饰或仅 dTdT 悬垂即可，预算花在多订 1–2 条候选上更划算；原代/免疫敏感细胞优先在个别位置加 2′-OMe 并避开免疫刺激基序；体内实验必须系统修饰并配合递送系统，超出入门范围。修饰解决稳定性和免疫问题，解决不了序列选错——敲低不好先回第 3、7 章。Decide by scenario: for in-vitro short-term work, standard unmodified or dTdT-overhang siRNA is enough — the budget is better spent on 1–2 extra candidates. For primary or immune-sensitive cells, add 2′-OMe at selected positions and avoid immunostimulatory motifs. In-vivo work needs systematic modification plus a delivery system, beyond beginner scope. Modifications fix stability and immunity problems; they cannot fix a wrongly chosen sequence — poor knockdown sends you back to Chapters 3 and 7 first.
第 6 章 · 验证：qPCR 与 WesternChapter 6 · Validation: qPCR and Western Blot
| 必备对照Essential control | 作用Purpose |
|---|
| 阴性对照 siRNA（NC）Non-targeting control (NC) | 定义基线，排除转染本身的影响Defines the baseline; excludes transfection effects |
| 阳性对照 siRNAPositive-control siRNA | 证明转染体系本身有效Proves the transfection system itself works |
| 仅转染试剂对照（mock）Mock (reagent only) | 区分试剂毒性与 siRNA 效应Separates reagent toxicity from siRNA effects |
| 未处理对照Untreated control | 细胞状态与基础表达参照Cell-state and baseline-expression reference |
qPCR 要点：转染后 24–48 h 收 RNA；引物不要与 siRNA 靶位点重叠（避免测到被切断未降解的片段），跨外显子连接处设计可排除基因组 DNA 干扰；内参至少 1 个经验证稳定的看家基因（靶基因本身是看家基因时必须更换）；用 ΔΔCt 法以阴性对照为校准，报告剩余表达百分比；生物学重复至少 3 次。判读：mRNA 剩余 ≤30% 算好，30%–50% 可用但要看蛋白与表型，>50% 先走第 7 章排障。qPCR points: harvest RNA 24–48 h after transfection; primers must not overlap the siRNA target site (you would measure cut-but-undegraded fragments), and exon-junction primers also exclude genomic DNA; use at least one validated stable reference gene (never the target itself if it is a housekeeping gene); quantify by ΔΔCt against the negative control and report remaining expression %; run at least 3 biological replicates. Interpretation: ≤ 30% mRNA remaining is good; 30–50% is usable pending protein and phenotype; above 50%, go to Chapter 7 before judging the sequence.
Western 要点：mRNA 降不等于蛋白降——长半衰期蛋白要到转染后 72–96 h 才看得出。48/72/96 h 分点收蛋白；等量上样，用总蛋白染色或看家蛋白归一；抗体先用过表达或已知阳性样本验证特异性与分子量；灰度定量归一后与阴性对照比，和 qPCR 结果并排报告。表型实验至少用 2 条独立 siRNA 重复，并保留 rescue 条件。Western points: mRNA down does not mean protein down — long-half-life proteins may not fall until 72–96 h. Harvest at 48/72/96 h; load equal amounts and normalize to total-protein stain or a housekeeping protein; validate the antibody first with overexpression or a known positive sample (specificity and molecular weight); quantify bands after normalization and report next to the qPCR result. Phenotype experiments need at least 2 independent siRNAs, with rescue kept possible.
| qPCR 与 Western 对不上When qPCR and Western disagree | 优先怀疑Suspect first |
|---|
| mRNA 大降，蛋白不降mRNA way down, protein unchanged | 蛋白半衰期长（延长时间点）；抗体问题；翻译补偿Long protein half-life (extend time points); antibody problem; translational compensation |
| mRNA 不降，蛋白降mRNA unchanged, protein down | qPCR 引物覆盖了未降解片段；引物测到了其他转录本qPCR primers sit on an undegraded fragment; primers detect another transcript |
| 两者都不降，阳性对照正常Neither drops; positive control fine | 这条序列本身无效，换下一条候选This sequence simply doesn't work — move to the next candidate |
| 两者都不降，阳性对照也不降Neither drops; positive control also fails | 转染体系问题，与序列无关A transfection-system problem, unrelated to the sequence |
第 7 章 · 常见问题排障Chapter 7 · Troubleshooting
| 现象Symptom | 先查Check first | 改法Fix |
|---|
| 全部候选都不敲低No candidate knocks down | 阳性对照；转染效率；细胞状态Positive control; transfection efficiency; cell state | 先优化转染（试剂量、10–50 nM 梯度、细胞密度），体系通了再评序列Optimize transfection first (reagent dose, 10–50 nM gradient, cell density); judge sequences only after the system works |
| 只有 1 条有效Only one works | 属正常分布，不是故障Normal distribution, not a fault | 有效条做主力，表型再补 1 条独立有效条印证Use the working one as primary; corroborate the phenotype with a second independent one |
| 敲低有，但细胞大量死亡Knockdown works but cells die heavily | mock 组与 NC 组是否也死Do mock and NC groups also die? | mock 死 = 试剂毒性，降剂量/换试剂；仅靶 siRNA 死 = 可能必需基因真表型或强脱靶，换 seed 不同的序列复核Mock dying = reagent toxicity: lower dose or switch reagent. Only target siRNA dying = possible essential-gene phenotype or strong off-target: verify with a different-seed sequence |
| 重复之间波动极大Huge variation between replicates | 铺板密度与收样时间是否一致Consistent seeding density and harvest times? | 固定密度与时间点；转染在对数生长期做Fix density and time points; transfect during log-phase growth |
| 敲低 48 h 好、后面反弹Good at 48 h, rebounds later | siRNA 被稀释/降解（分裂快的细胞明显）siRNA diluted/degraded (worst in fast-dividing cells) | 缩短实验窗、二次转染，或改用 shRNAShorten the window, re-transfect, or switch to shRNA |
| 表型与敲低程度对不上Phenotype doesn't track knockdown | 是否只有高脱靶风险的那条有表型Does only the high-off-target-risk siRNA show the phenotype? | 多条独立 siRNA 交叉验证 + rescue；对不上的按脱靶处理Cross-validate with independent siRNAs + rescue; treat unmatched phenotypes as off-target |
排障顺序口诀：先对照，后体系，再序列。阳性对照正常、转染效率达标之后，才有资格说某条序列不好。Order of operations: controls first, then the system, then the sequence. Only when the positive control works and transfection efficiency passes do you get to call a sequence bad.
第 8 章 · 一个完整设计实例（教学演示）Chapter 8 · A Complete Worked Example (Teaching Demonstration)
先说清：本章演示序列是按规则构造的教学窗口，不对应任何已验证的商品 siRNA。真实项目从你的目标转录本取窗之后，其余步骤与本章完全相同，订购前必须按最新版 RefSeq 重新核对并完成第 4 章检查。Up front: the sequences in this chapter are teaching windows constructed by the rules; they are not any validated commercial siRNA. In a real project, once you take windows from your actual target transcript, every later step is identical to this chapter — and before ordering you must re-verify against the current RefSeq and complete the Chapter 4 checks.
任务：在人源细胞系敲低某基因全部主要转录本，做 72 h 内表型实验，体外脂质体转染，预算够 3 条候选 + 阴性 + 阳性对照。步骤：查 RefSeq 记录主转录本编号版本并下载 CDS → 查 dbSNP 标出共有区常见 SNP → 共有 CDS 前/中/后分散取窗 → 19 nt 滑窗过硬规则初筛，进入评分的 3 条窗口如下：Task: knock down all major transcripts of a gene in a human cell line for a 72 h phenotype experiment, lipid transfection in vitro, budget for 3 candidates + negative + positive controls. Steps: look up the main transcript in RefSeq (record ID and version, download the CDS) → mark common SNPs in the shared region from dbSNP → take windows spread across the shared CDS → slide 19 nt windows through the hard rules. Three windows reaching scoring:
| 候选Candidate | 靶序列（19 nt，正义链方向）Target sequence (19 nt, sense) | GC | guide 链Guide strand | seed (2–8) | 得分 / 决定Score / decision |
|---|
| 1 | GUAUGACAACAGCCUCAAG | 47.4% | CUUGAGGCUGUUGUCAUAC | UUGAGGC | 5/9 · guide 5′ 端为 C、不对称性差，降级备用5/9 · guide starts with C, poor asymmetry — demoted to backup |
| 2 | GCAAUGACAUGCAUUCAUU | 36.8% | AAUGAAUGCAUGUCAUUGC | AUGAAUG | 9/9 · 不对称性最好，seed GC 仅 28.6%，首选下单9/9 · best asymmetry, seed GC only 28.6% — first choice to order |
| 3 | GGAAUUCAUGCAGACUUCA | 42.1% | UGAAGUCUGCAUGAAUUCC | GAAGUCU | 7/9 · 第 10、13 位失分，可作独立第二条，下单7/9 · loses points at positions 10 and 13; a valid independent second — order |
之后对候选 2、3 做全长 BLAST + seed 匹配谱检查（两条 seed 完全不同，符合互相印证要求），再补第 3 条分散位置候选凑满 3 条。下单规格：19 nt 双链 + 3′ dTdT 悬垂，guide 链 5′ 磷酸状态向厂商确认，阴性与阳性对照一并订购。验证排程：荧光 siRNA 优化转染 → 48 h qPCR → 72/96 h Western。本例要点不在某条序列，而在顺序：选区 → 硬规则 → 算 guide → 评分排序 → 脱靶双查 → 才下单。Then run full-length BLAST + seed-profile checks on candidates 2 and 3 (their seeds are completely different, satisfying the cross-validation requirement) and add a third spread-out candidate to make three. Order spec: 19 nt duplex + 3′ dTdT overhang, guide 5′ phosphate status confirmed with the vendor, negative and positive controls ordered together. Validation schedule: fluorescent-siRNA transfection optimization → qPCR at 48 h → Western at 72/96 h. The lesson is the order, not any one sequence: region → hard rules → compute the guide → score and rank → dual off-target checks → only then order.
附录 · 订购前总清单与参数速查Appendix · Pre-Order Checklist and Quick Parameters
- 目标转录本 ID + 版本号已记录，靶区与"敲全部亚型 / 单一亚型"的目标一致Target transcript ID + version recorded; target region matches the all-isoform vs single-isoform goal
- 靶区避开常见 SNP、低复杂度区与同源基因保守区Target region avoids common SNPs, low-complexity stretches, and homolog-conserved regions
- 候选共 3–4 条，位置分散，每条 GC 在 30%–52%3–4 candidates, positions spread out, each with GC 30%–52%
- 每条已算出 guide 链，guide 5′ 端为 A/U，不对称性成立Guide strand computed for each; guide 5′ end is A/U; asymmetry holds
- 每条已按 9 项评分表打分排序，没有只凭单一工具下单Each scored and ranked on the 9-item table; no order placed on a single tool's say-so
- 全长 BLAST 已做（短序列模式），无 15 nt 以上连续匹配的非目标转录本Full-length BLAST done (short-sequence mode); no non-target transcript with a ≥ 15 nt contiguous match
- seed（guide 第 2–8 位）匹配谱已查，几条候选 seed 互不相同Seed (guide positions 2–8) profile checked; candidate seeds all differ
- 下单规格写明：长度、dTdT 悬垂、5′ 磷酸状态、修饰及位点Order spec states length, dTdT overhang, 5′ phosphate status, and any modifications with positions
- 阴性对照、阳性对照 siRNA 一并订购Negative- and positive-control siRNAs ordered together
- 验证方案已定：qPCR 引物不覆盖靶位点，Western 时间梯度与抗体已验证Validation planned: qPCR primers clear of the target site; Western time course and antibody validated
| 参数Parameter | 常用值Typical value |
|---|
| 双链区长度Duplex length | 19 nt (+ 2 nt 3′ 悬垂overhang) |
| GC 含量窗口GC window | 30%–52% |
| 转染终浓度（优化起点）Transfection concentration (optimization start) | 10–50 nM 梯度gradient |
| qPCR 收样qPCR harvest | 转染后24–48 h after transfection |
| Western 收样Western harvest | 48 / 72 / 96 h |
| 敲低达标线（mRNA）Knockdown pass line (mRNA) | 剩余 ≤30%，即敲低 ≥70%≤ 30% remaining, i.e. ≥ 70% knockdown |
| 表型证据标准Phenotype evidence standard | ≥2 条独立 siRNA 同表型 + rescue≥ 2 independent siRNAs, same phenotype + rescue |
主要规则来源：Reynolds et al., Nature Biotechnology 2004；Ui-Tei et al., Nucleic Acids Research 2004；Schwarz et al., Cell 2003 与 Khvorova et al., Cell 2003（热力学不对称性）；Elbashir et al., Nature 2001。数据库与工具：NCBI RefSeq、BLAST、dbSNP、Ensembl。本指南为实验方法教育资料，不构成对任何具体产品或厂商的推荐；演示序列为教学构造，订购与实验前请按最新数据库与所在实验室安全规范复核。Main rule sources: Reynolds et al., Nature Biotechnology 2004; Ui-Tei et al., Nucleic Acids Research 2004; Schwarz et al., Cell 2003 and Khvorova et al., Cell 2003 (thermodynamic asymmetry); Elbashir et al., Nature 2001. Databases and tools: NCBI RefSeq, BLAST, dbSNP, Ensembl. This guide is educational material on experimental method and recommends no specific product or vendor; the demonstration sequences are teaching constructs — re-verify against current databases and your lab's safety rules before ordering or experimenting.
lncRNA 设计指南lncRNA Design Guide
从定位判断、策略选择（siRNA / ASO / CRISPRi / 敲除）、序列设计到验证——第一次靶向长链非编码 RNA 就能走完的完整流程。本书解决的问题：lncRNA 不是"不会翻译的 mRNA"。它丰度低、亚型多、很多待在细胞核里，用敲 mRNA 的 siRNA 思路直接套，常常一条都敲不动。这本书先帮你选对工具，再谈序列，每一步都给判断标准和清单。From localization to strategy (siRNA / ASO / CRISPRi / knockout), sequence design, and validation — the complete workflow for your first lncRNA targeting project. The problem this guide solves: an lncRNA is not an "mRNA that doesn't translate." It is low-abundance, multi-isoform, and often stuck in the nucleus — apply the mRNA-knockdown siRNA playbook directly and most of the time nothing knocks down. This guide picks the right tool first and talks sequence second, with a decision standard and a checklist at every step.
适用对象：第一次独立设计 lncRNA 敲低 / 抑制实验的实验人员（细胞实验为主）Audience: first-time lncRNA knockdown / repression designers (mainly cell culture)
前置要求：会查 RefSeq / Ensembl，会做基本的转染和 qPCR；CRISPR 部分需有慢病毒或电转条件Prerequisites: can use RefSeq / Ensembl; basic transfection and qPCR; lentivirus or electroporation access for the CRISPR parts
版本 v1.0 · 2026 年 11 月Version 1.0 · November 2026
与 siRNA 指南最大的不同：敲 mRNA 时工具默认是 siRNA；敲 lncRNA 时没有默认工具。先查定位、再选工具，是本书的第一条纪律。The biggest difference from the siRNA guide: knocking down mRNA defaults to siRNA; knocking down lncRNA has no default tool. Check localization first, then pick the tool — the first discipline of this guide.
使用方法：先读第 1–2 章再动手——定位决定用 ASO 还是 siRNA，策略选错后面全是白费。每个项目至少准备 2 种独立试剂（如 2 条 ASO，或 ASO + CRISPRi），表型要互相印证。第 5 章的脱靶双重检查和最后的订购前清单是硬门槛，没过不要下单。没敲下来之前不要下结论说这个 lncRNA"没功能"——先看第 8 章排障，十有八九是定位与工具不匹配。How to use this guide: read Chapters 1–2 before touching anything — localization decides ASO vs siRNA, and a wrong strategy wastes everything after it. Prepare at least 2 independent reagents per project (e.g. 2 ASOs, or ASO + CRISPRi) so phenotypes corroborate each other. The dual off-target checks in Chapter 5 and the final pre-order checklist are hard gates — do not order until you pass them. Before concluding an lncRNA "has no function" because it won't knock down, check Chapter 8: nine times out of ten it is a localization–tool mismatch.
第 1 章 · lncRNA 生物学基础与分类Chapter 1 · lncRNA Biology and Classification
长链非编码 RNA 指长度大于 200 nt、通常不翻译成蛋白质的转录本。多数 lncRNA 像 mRNA 一样有 5′ 帽和 poly(A) 尾、也经剪接，但有三点和 mRNA 明显不同，每一点都直接影响实验设计。丰度低。多数 lncRNA 的拷贝数只有常见 mRNA 的十分之一甚至更低。后果：qPCR 的 Ct 值偏高、检测波动大，敲低百分比要用足生物学重复才可信；单细胞和低起始量样本尤其要小心。组织与细胞特异性强。lncRNA 的表达谱比 mRNA 窄得多，换一个细胞系可能根本不表达。后果：动手设计之前，先用 RNA-seq 数据或 qPCR 确认你的细胞里它真的表达，以及表达的是哪个亚型。定位决定功能，也决定可及性。许多 lncRNA 滞留在细胞核内参与染色质调控，而 RISC / siRNA 机器主要在细胞质工作。后果：核内 lncRNA 用 siRNA 经常敲不动，要用能在核内工作的 ASO（RNase H1 在核内有活性）或从转录层面下手的 CRISPRi（第 2 章）。Long non-coding RNAs are transcripts longer than 200 nt that are usually not translated into protein. Most lncRNAs carry a 5′ cap and a poly(A) tail and are spliced, like mRNA — but they differ from mRNA in three ways, and each one directly changes experimental design. Low abundance. Most lncRNAs are present at a tenth or less of the copy number of a typical mRNA. Consequence: qPCR Ct values run high and noisy, so knockdown percentages are only believable with enough biological replicates — be extra careful with single-cell and low-input samples. Strong tissue and cell specificity. lncRNA expression profiles are far narrower than mRNA; switch cell lines and it may not be expressed at all. Consequence: before designing anything, confirm with RNA-seq or qPCR that your cells actually express it, and which isoform. Localization decides function — and accessibility. Many lncRNAs stay in the nucleus doing chromatin regulation, while the RISC / siRNA machinery works mainly in the cytoplasm. Consequence: nuclear lncRNAs often resist siRNA; use an ASO that works in the nucleus (RNase H1 is active there) or CRISPRi, which acts at the transcription level (Chapter 2).
| 功能方式Mode of action | 机制Mechanism | 敲低后该看什么What to read out after knockdown |
|---|
| 信号 / 诱饵Signal / decoy | 结合并隔离转录因子或 miRNABinds and sequesters transcription factors or miRNAs | 被隔离因子的靶基因表达、游离 miRNA 的靶 mRNATarget-gene expression of the sequestered factor; target mRNAs of the freed miRNA |
| 支架Scaffold | 同时结合多个蛋白复合体，把它们组装到一起Binds several protein complexes at once and assembles them | 复合体成员的互作与下游通路活性Interaction of complex members and downstream pathway activity |
| 向导Guide | 把染色质修饰复合体带到特定基因组位点Recruits chromatin-modifying complexes to specific genomic loci | 目标位点的组蛋白修饰与该基因表达Histone marks at the target locus and that gene's expression |
| 顺式调控Cis regulation | 在转录地附近调控邻近基因Regulates neighboring genes near its site of transcription | 同一位点邻近基因的表达（注意与位点效应区分，第 2 章）Expression of genes neighboring the locus (distinguish from locus effects, Chapter 2) |
先分清 cis 还是 trans：只影响邻近基因的 lncRNA 多为 cis 作用，它的"功能"可能来自转录过程或 DNA 元件本身，而不只是 RNA 分子——这决定了你能不能用外源回补（rescue）以及怎么解释 CRISPR 结果。Sort out cis vs trans first: an lncRNA that affects only neighboring genes usually acts in cis, and its "function" may come from the act of transcription or the DNA element itself, not just the RNA molecule — which decides whether an exogenous rescue is possible and how to interpret CRISPR results.
| 按基因组位置分类Classification by genomic position | 位置Position | 设计注意Design notes |
|---|
| 基因间 lncRNA（lincRNA）Intergenic lncRNA (lincRNA) | 两个蛋白编码基因之间，独立转录Between two protein-coding genes, independently transcribed | 邻居干扰最小，CRISPR 操作相对干净Least neighbor interference; cleanest for CRISPR work |
| 反义 lncRNA（antisense / NAT）Antisense lncRNA (NAT) | 与蛋白编码基因反义链重叠Overlaps a protein-coding gene on the opposite strand | 试剂和 qPCR 引物必须链特异；敲它可能顺带影响正义基因Reagents and qPCR primers must be strand-specific; knocking it down may also affect the sense gene |
| 内含子 lncRNA（intronic）Intronic lncRNA | 藏在某基因的内含子中Inside an intron of another gene | 避开宿主基因的剪接位点；测表达要排除未剪接前体干扰Avoid the host gene's splice sites; exclude unspliced precursor when measuring expression |
| 正义重叠 lncRNA（sense-overlapping）Sense-overlapping lncRNA | 与编码基因同链重叠、共享外显子序列Same-strand overlap with a coding gene, sharing exon sequence | 共享区段不能选靶，否则等于敲编码基因；只选它独有区段Never target the shared segment or you knock down the coding gene; use only its unique regions |
| 增强子 RNA（eRNA）等短寿命类型Enhancer RNA (eRNA) and other short-lived types | 增强子区转录、半衰期极短、无 poly(A) 的多Transcribed from enhancers, very short half-life, often no poly(A) | 一般不做常规敲低对象；检测需随机引物逆转录（第 7 章）Usually not a routine knockdown target; detection needs random-primed reverse transcription (Chapter 7) |
| 按亚细胞定位分类Classification by subcellular localization | 典型机制Typical mechanisms | 首选工具（第 2 章展开）First-choice tool (Chapter 2) |
|---|
| 核内为主Mainly nuclear | 染色质调控、核内结构支架、剪接调控Chromatin regulation, nuclear structural scaffolds, splicing regulation | ASO gapmer 或 CRISPRi；siRNA 成功率低ASO gapmer or CRISPRi; siRNA rarely works |
| 胞质为主Mainly cytoplasmic | miRNA 海绵（ceRNA）、翻译调控、mRNA 稳定性miRNA sponge (ceRNA), translational control, mRNA stability | siRNA 可用；ASO 同样有效siRNA works; ASO works too |
| 核质双分布Both nucleus and cytoplasm | 穿梭型，两种机制并存Shuttling; both mechanisms coexist | ASO 优先（核质通吃），siRNA 只敲得动胞质部分ASO first (works in both compartments); siRNA only knocks down the cytoplasmic fraction |
定位怎么查：先查数据库与文献（lncATLAS 等有相对浓度指数），再用自己的细胞做核质分离 qPCR 或 RNA FISH 实测确认——定位随细胞类型变化，文献值只能当线索，不能当结论。How to check localization: start with databases and the literature (lncATLAS and similar resources report relative concentration indices), then confirm in your own cells with nuclear/cytoplasmic fractionation qPCR or RNA FISH — localization varies by cell type, so literature values are clues, not conclusions.
分类不是背概念，是排雷：反义和重叠型的"邻居问题"、核内型的"工具问题"，是 lncRNA 项目翻车率最高的两个坑，后面每一章都在回扣这两点。Classification is not memorization — it is mine-clearing: the "neighbor problem" of antisense and overlapping types and the "tool problem" of nuclear types are the two most common ways lncRNA projects fail, and every later chapter comes back to them.
第 2 章 · 靶向策略：四种工具怎么选Chapter 2 · Targeting Strategies: Choosing Among Four Tools
四种工具作用在不同层面：siRNA 和 ASO 打 RNA 分子，CRISPRi 压转录，CRISPR 敲除改 DNA。先看对比，再按定位做决定。The four tools act at different levels: siRNA and ASO hit the RNA molecule, CRISPRi represses transcription, and CRISPR knockout changes the DNA. Compare them first, then decide by localization.
| 工具Tool | 机制Mechanism | 核内靶Nuclear targets | 优缺点要点Key pros and cons |
|---|
| siRNA | RISC / Ago2 切割（主要在胞质）RISC / Ago2 cleavage (mainly cytoplasmic) | 差Poor | 便宜快、规则成熟；核内 lncRNA 常无效；瞬时敲低Cheap, fast, mature rules; usually ineffective on nuclear lncRNAs; transient knockdown |
| ASO gapmerASO gapmer | 中央 DNA gap 招募 RNase H1 切割 RNA（核质均可）Central DNA gap recruits RNase H1 to cut RNA (nucleus and cytoplasm) | 好Good | lncRNA 敲低的主力；需化学修饰（第 6 章）；单链、可裸转染（gymnosis）或脂质体递送The workhorse for lncRNA knockdown; needs chemical modification (Chapter 6); single-stranded, deliverable by gymnosis or lipofection |
| CRISPRi | dCas9-KRAB 结合启动子区，压制转录起始dCas9-KRAB binds the promoter and represses transcription initiation | 好（源头压制）Good (acts at the source) | 不依赖 RNA 定位；需稳定表达 dCas9；只对准确定位的 TSS 有效Independent of RNA localization; needs stable dCas9 expression; works only for a precisely mapped TSS |
| CRISPR 敲除CRISPR knockout | 删除启动子 / 关键外显子或插入 poly(A) 终止信号Deletes the promoter / key exon or inserts a poly(A) termination signal | 好（永久）Good (permanent) | 永久丢失适合建系；但改动 DNA，表型可能来自 DNA 元件或邻居基因，不等于 RNA 功能丢失Permanent loss suits stable lines — but the DNA change means the phenotype may come from a DNA element or a neighbor gene, not from loss of the RNA's function |
决策顺序（照这个顺序问四个问题）：1. 它在哪？核内为主 → ASO 或 CRISPRi 起步；胞质为主 → siRNA 与 ASO 都可，先试便宜的。2. 要多久？短期表型（几天）→ siRNA / ASO；长期或建系 → CRISPRi 稳定株或敲除。3. 位点干不干净？与邻居基因重叠或共享调控元件 → 避免 DNA 层面操作（敲除、甚至 CRISPRi），只用 RNA 层面工具。4. 结论要多硬？要发表级证据 → 至少两种原理不同的工具给出同一表型（如 ASO + CRISPRi）。Decision order (ask these four questions in sequence): 1. Where is it? Mainly nuclear → start with ASO or CRISPRi; mainly cytoplasmic → siRNA and ASO both work, try the cheaper one first. 2. How long? Short-term phenotype (days) → siRNA / ASO; long-term or stable lines → CRISPRi stable line or knockout. 3. Is the locus clean? Overlap with a neighbor gene or shared regulatory elements → avoid DNA-level operations (knockout, even CRISPRi) and use RNA-level tools only. 4. How strong must the conclusion be? Publication-grade evidence → at least two mechanistically different tools giving the same phenotype (e.g. ASO + CRISPRi).
位点效应（locus effect）是 CRISPR 路线最大的解释陷阱：删掉一段 DNA 后出现的表型，可能来自被删的增强子、被波及的邻居基因，而不是 lncRNA 分子本身。CRISPR 结果必须有 RNA 层面工具（ASO / siRNA）的结果对照，才能归因到 RNA。The locus effect is the biggest interpretation trap of the CRISPR route: a phenotype after deleting a DNA stretch may come from a deleted enhancer or an affected neighbor gene, not from the lncRNA molecule itself. CRISPR results must be controlled with results from an RNA-level tool (ASO / siRNA) before attributing anything to the RNA.
第 3 章 · 靶点选择：先定转录本，再定区域Chapter 3 · Target Selection: Fix the Transcript First, Then the Region
lncRNA 的亚型（isoform）往往比 mRNA 更复杂，同一位点不同转录本可以差出几个外显子。选区顺序错了，后面规则再对也是敲错对象。第一步：确认表达与转录本。在 RefSeq / Ensembl（或 GENCODE，lncRNA 注释最全）里列出该基因的全部转录本，在你的细胞里用 RNA-seq 或多扩增子 qPCR 确认实际表达的是哪几条。想敲掉全部亚型：靶点落在所有表达转录本共有的组成型外显子上；只想敲某一个亚型：选它独有的外显子或剪接连接处。记录转录本 ID 与版本号；lncRNA 注释更新频繁，去年下的序列今年可能已改版。第三步：CRISPR 路线的选区是另一套。CRISPRi 的靶点不在转录本上，在启动子上：sgRNA 设计在 TSS 上下游窗口内（常用 TSS 下游 -50 至 +300 bp 一带活性最好），所以 TSS 定位必须准确（查 GENCODE 注释或 CAGE 数据）。敲除则优先删启动子 + 第一个外显子，或插入 poly(A) 终止信号，大片段删外显子前先核对该区段是否含邻居基因的调控元件。lncRNA isoforms are often more complex than mRNA — transcripts from one locus can differ by several exons. Get the selection order wrong and even perfect rules knock down the wrong thing. Step 1: confirm expression and transcripts. List every transcript of the gene in RefSeq / Ensembl (or GENCODE, the most complete lncRNA annotation) and confirm which ones are actually expressed in your cells by RNA-seq or multi-amplicon qPCR. To knock down all isoforms, target a constitutive exon shared by every expressed transcript; to hit one isoform only, pick its unique exon or splice junction. Record transcript IDs and versions — lncRNA annotation updates often, and last year's sequence may have changed. Step 3: CRISPR routes pick regions differently. A CRISPRi target is not on the transcript but on the promoter: design sgRNAs in the window around the TSS (usually −50 to +300 bp downstream of the TSS works best), so the TSS must be mapped accurately (GENCODE annotation or CAGE data). For knockout, prefer deleting the promoter + first exon, or inserting a poly(A) termination signal; before deleting a large exon stretch, check whether it contains regulatory elements of a neighbor gene.
| 第二步：避开这些位置Step 2: avoid these positions | 为什么Why |
|---|
| 重复元件区（Alu、LINE 等）Repeat elements (Alu, LINE, etc.) | lncRNA 外显子常含转座子来源序列，在全转录组有成百上千个近似匹配，脱靶不可控lncRNA exons often carry transposon-derived sequence with hundreds of near-matches across the transcriptome — off-targets become uncontrollable |
| 与对侧基因重叠的区段Segments overlapping the opposite-strand gene | 反义重叠区的一条 ASO / siRNA 可能同时动两个基因，表型无法归因One ASO / siRNA in an antisense overlap can hit both genes, and the phenotype cannot be attributed |
| 已知 SNP 位点Known SNP sites | 你的细胞系若是另一等位基因，错配直接让试剂失效；查 dbSNP 后绕开If your cell line carries the other allele, the mismatch kills the reagent; check dbSNP and route around |
| 剪接位点与分支点附近（ASO）Near splice sites and branch points (ASO) | 除非目的就是调控剪接，否则避开外显子两端约 20 ntUnless splicing regulation is the goal, stay ~20 nt away from exon ends |
| 强二级结构 / 高 GC 折叠区Strong secondary structure / high-GC folds | lncRNA 结构域多，靶位被自身结构包裹时 ASO 与 siRNA 都够不着；有条件时用结构预测或已发表的可及性数据选开放区lncRNA is rich in structure; when the target site is buried in its own fold, neither ASO nor siRNA can reach it — use structure prediction or published accessibility data to pick open regions when available |
- 实际表达的转录本已在目标细胞中确认，ID 与版本已记录Actually expressed transcripts confirmed in the target cells; IDs and versions recorded
- 靶区在组成型外显子上（或已明确说明只敲单一亚型的理由）Target region is on a constitutive exon (or the single-isoform rationale is stated)
- 靶区不含重复元件、无常见 SNP、不与对侧基因重叠Target region has no repeat elements, no common SNPs, no overlap with the opposite-strand gene
- CRISPRi 项目：TSS 已核准，sgRNA 窗口已定CRISPRi projects: TSS verified, sgRNA window set
第 4 章 · 设计规则：三种试剂，三套规则Chapter 4 · Design Rules: Three Reagents, Three Rule Sets
siRNA（胞质靶或双分布靶的胞质部分）：沿用 siRNA 通用规则：19 nt 双链区 + 3′ dTdT 悬垂，GC 30%–52%，guide 链 5′ 端为 A/U 且该端热力学不稳定（不对称性），详见《siRNA 设计指南》第 3 章。lncRNA 额外一条：避开多亚型间不共有的区段和重复元件区（第 3 章），这两条比单条序列的分数更重要。CRISPRi sgRNA（转录层面）：在 TSS -50 至 +300 bp 窗口内设计 3–5 条 sgRNA 一起筛，这个窗口对 KRAB 压制最敏感；窗口外效率掉得很快。沿用 sgRNA 通用规则：20 nt spacer，GC 40%–60%，避开 4 连 T（Pol III 终止信号），查全基因组脱靶（第 5 章）。多个 TSS 的基因要分别覆盖每个实际使用的 TSS，否则压不全。siRNA (cytoplasmic targets, or the cytoplasmic fraction of dual-localized targets): follow the general siRNA rules — 19 nt duplex + 3′ dTdT overhang, GC 30%–52%, guide 5′ end A/U with that end thermodynamically unstable (asymmetry); see Chapter 3 of the siRNA Design Guide. One extra lncRNA rule: avoid regions not shared across isoforms and repeat-element regions (Chapter 3) — these two matter more than any single sequence's score. CRISPRi sgRNA (transcription level): design 3–5 sgRNAs inside the TSS −50 to +300 bp window and screen them together; this window is most sensitive to KRAB repression and efficiency drops fast outside it. Follow the general sgRNA rules: 20 nt spacer, GC 40%–60%, avoid 4 consecutive Ts (a Pol III termination signal), and check genome-wide off-targets (Chapter 5). Genes with multiple TSSs need each actually used TSS covered, or repression will be incomplete.
| ASO gapmer 规则ASO gapmer rule | 标准与原因Standard and reason |
|---|
| 总长 16–20 ntTotal length 16–20 nt | 短于 16 nt 特异性不够；长于 20 nt 摄取与脱靶都变差Shorter than 16 nt: too little specificity. Longer than 20 nt: worse uptake and more off-targets |
| 中央 DNA gap 8–10 ntCentral DNA gap 8–10 nt | gap 是 RNase H1 的切割区，太短招募不了 RNase H1；两侧为 2′ 修饰翼（wing），常见 5-10-5 或 3-10-3 构型The gap is the RNase H1 cleavage zone; too short and RNase H1 is not recruited. Both sides are 2′-modified wings; common layouts are 5-10-5 or 3-10-3 |
| GC 含量 40%–60%GC content 40%–60% | ASO 靠高亲和力结合，窗口比 siRNA 略高；过高易自聚集与非特异结合ASOs bind through high affinity, so the window sits slightly higher than siRNA; too high and self-aggregation and non-specific binding rise |
| 避开 GGGG 等 G 四联体倾向序列Avoid GGGG and other G-quadruplex-prone runs | 易形成 G-quadruplex，降低有效浓度并增加非特异蛋白结合Prone to G-quadruplex formation, lowering effective concentration and raising non-specific protein binding |
| 避开强免疫刺激基序Avoid strong immunostimulatory motifs | 含未甲基化 CpG 等基序可激活 TLR9 样反应，引起与靶无关的表型Motifs such as unmethylated CpG can trigger TLR9-like responses and target-independent phenotypes |
| 序列内避免 ≥4 连排同碱基No runs of ≥ 4 identical bases | 合成质量和特异性都受影响Hurts both synthesis quality and specificity |
| # | 评分项（每满足一项记 1 分）Scoring item (1 point each) | 适用Applies to |
|---|
| 1 | 靶区在组成型外显子内，且在所有目标转录本中一致Target region is in a constitutive exon, identical across all target transcripts | siRNA / ASO |
| 2 | 靶区不含重复元件、无常见 SNPTarget region has no repeat elements and no common SNPs | siRNA / ASO |
| 3 | GC 在目标窗口内（siRNA 30%–52%；ASO 40%–60%）GC inside the target window (siRNA 30%–52%; ASO 40%–60%) | 两者Both |
| 4 | 无 4 个以上同碱基连排、无 GGGGNo runs of 4+ identical bases, no GGGG | 两者Both |
| 5 | 靶区不与对侧链基因或其他转录本重叠Target region does not overlap the opposite-strand gene or other transcripts | 两者Both |
| 6 | 预测二级结构为相对开放区（或有可及性数据支持）Predicted secondary structure is relatively open (or accessibility data supports it) | 两者Both |
| 7 | siRNA：guide 链 5′ 端为 A/U，不对称性成立siRNA: guide 5′ end is A/U and asymmetry holds | siRNA |
| 8 | ASO：gap 8–10 nt、总长 16–20 nt、构型明确（如 5-10-5）ASO: gap 8–10 nt, total length 16–20 nt, defined layout (e.g. 5-10-5) | ASO |
| 9 | seed 区（siRNA）或全长近错配谱（ASO）相对干净（第 5 章）Seed profile (siRNA) or near-mismatch spectrum (ASO) relatively clean (Chapter 5) | 两者Both |
| 10 | 与同项目其他候选位置分散、序列互不重叠Positions spread out from other candidates in the project; sequences do not overlap | 两者 / sgRNABoth / sgRNA |
怎么用分数：8 分以上优先下单，6–7 分可作独立第二、第三条，6 分以下降级备用。分数只做排序——lncRNA 没有任何评分能保证单条必中，这正是每个项目要备 3 条候选、且最好跨两种工具的原因。How to use the score: 8+ points: order first; 6–7: usable as independent second or third reagents; below 6: demote to backup. The score only ranks — no scoring system can guarantee a single lncRNA reagent will work, which is why every project prepares 3 candidates, preferably across two tool types.
| 场景Scenario | 起订组合Recommended starting set |
|---|
| 核内 lncRNA、短期实验Nuclear lncRNA, short-term experiment | ASO gapmer 3 条 + 阴性对照 ASO（同化学、打乱序列）3 ASO gapmers + control ASO (same chemistry, scrambled sequence) |
| 胞质 lncRNA、短期实验Cytoplasmic lncRNA, short-term experiment | siRNA 3 条（或 ASO 3 条）+ 阴性对照3 siRNAs (or 3 ASOs) + negative control |
| 长期 / 建系Long-term / stable lines | CRISPRi sgRNA 4–5 条（覆盖 TSS 窗口）+ 非靶向 sgRNA 对照4–5 CRISPRi sgRNAs (covering the TSS window) + non-targeting sgRNA control |
| 发表级结论Publication-grade conclusion | 上面任一组合 + 另一种原理的工具复现表型Any set above + a second, mechanistically different tool reproducing the phenotype |
阴性对照要同化学：ASO 的对照必须是同样骨架与修饰的打乱序列，siRNA 的对照用非靶向 duplex。拿未修饰的 oligo 给修饰 ASO 当对照，测出来的差异分不清是序列还是化学。Controls must match chemistry: an ASO control must be a scrambled sequence with the same backbone and modifications; a siRNA control is a non-targeting duplex. Using an unmodified oligo as the control for a modified ASO makes it impossible to tell whether a difference comes from sequence or chemistry.
第 5 章 · 脱靶检查：全长 + seed 双重检查Chapter 5 · Off-Target Checking: Full-Length Plus Seed — Both Layers
脱靶检查分两层，查的是两种不同的风险，只做一层就宣布"无脱靶"是新手最常见的错误。CRISPR 试剂另有基因组层面的检查，并列在本页末尾。第一层：全长 BLAST（近全长匹配脱靶）。把候选的靶序列在 NCBI BLAST 中对目标物种 transcript 数据库做 blastn，选短序列模式（word size 调小至 7，关低复杂度过滤）。判读：除目标 lncRNA 外，若其他转录本有 15 nt 以上连续匹配，或总错配 ≤2，这条候选淘汰或降级。特别核对基因家族成员、假基因，以及与你靶区重叠的对侧链转录本。ASO 额外一条：RNase H1 能容忍少量错配继续切割，所以要把"错配 1–3 个"的近匹配位点也逐个列出来评估，而不是只看完全匹配。第二层：seed 区检查（miRNA 样脱靶，siRNA 必查）。取 guide 链第 2–8 位共 7 nt 的 seed，查它在全转录组 3′UTR 中的匹配频率；匹配越广，潜在被压制的无关基因越多。用带 seed 分析的工具或脱靶预测数据库复核；seed 的 GC 过高时脱靶风险额外升高。完全避开 seed 脱靶不现实，目标是选 seed 匹配谱相对干净、且同项目几条候选的 seed 互不相同的组合。CRISPR 试剂的检查（基因组层面）：sgRNA 用 CRISPOR、Cas-OFFinder 类工具查全基因组潜在脱靶位点，优先淘汰在外显子区有 ≤2 错配位点的 spacer。敲除用的双 sgRNA 还要核对删除区段的基因组内容：有没有邻居基因的启动子、增强子或 CTCF 位点落在里面（位点效应排雷）。Off-target checking has two layers covering two different risks; running only one layer and declaring "no off-targets" is the most common beginner mistake. CRISPR reagents need an additional genome-level check, listed at the end of this chapter. Layer 1: full-length BLAST (near-full-length off-targets). BLAST the candidate target sequence against the species' transcript database in NCBI BLAST using blastn short-sequence mode (word size down to 7, low-complexity filter off). Interpretation: apart from the target lncRNA, any other transcript with a ≥ 15 nt contiguous match, or ≤ 2 total mismatches, gets the candidate dropped or demoted. Check gene-family members, pseudogenes, and opposite-strand transcripts overlapping your target region especially. One extra ASO rule: RNase H1 tolerates a few mismatches and keeps cutting, so near-match sites with 1–3 mismatches must also be listed and evaluated one by one, not just perfect matches. Layer 2: seed check (miRNA-like off-targets — mandatory for siRNA). Take the 7 nt seed (guide positions 2–8) and check its match frequency across transcriptome 3′UTRs; the broader the matches, the more unrelated genes may be repressed. Re-check with a seed-analysis tool or off-target prediction database; a GC-rich seed raises the risk further. Avoiding seed off-targets entirely is unrealistic — aim for a relatively clean seed profile and candidates whose seeds all differ. CRISPR reagent check (genome level): search the whole genome for potential off-target sites with tools like CRISPOR or Cas-OFFinder, and drop spacers with ≤ 2-mismatch sites in exons first. For knockout pairs, also audit the deleted segment's genomic content: does it contain a neighbor gene's promoter, enhancer, or CTCF site (clearing locus-effect mines)?
- 全长 BLAST 已做（短序列模式），结果存档；无 15 nt 以上连续匹配的非目标转录本Full-length BLAST done (short-sequence mode), results archived; no non-target transcript with a ≥ 15 nt contiguous match
- seed（guide 第 2–8 位）匹配谱已查，同项目候选 seed 互不相同Seed (guide positions 2–8) profile checked; candidate seeds in the project all differ
- ASO 候选的 1–3 错配近匹配位点已列出并逐个评估ASO candidates' 1–3-mismatch near-match sites listed and evaluated one by one
- sgRNA（如用）全基因组脱靶已查，删除区段内容已核对sgRNA (if used) genome-wide off-targets checked; deleted segment contents verified
第 6 章 · 化学修饰：ASO 与 siRNA 是两套逻辑Chapter 6 · Chemical Modifications: ASO and siRNA Follow Different Logic
单链 ASO 裸奔进细胞会被核酸酶迅速降解，所以修饰对 ASO 是"必需品"；双链 siRNA 有蛋白保护，修饰多是"优化项"。把两套逻辑分开记。A naked single-stranded ASO is rapidly degraded by nucleases inside the cell, so modification is a necessity for ASO; a double-stranded siRNA is protected by protein, so modification is mostly an optimization. Keep the two logics separate.
| 修饰Modification | 作用Purpose | 注意Caution |
|---|
| 硫代磷酸酯（PS）骨架Phosphorothioate (PS) backbone | 抗核酸酶、促进蛋白结合与细胞摄取，ASO 全长标配Nuclease resistance, protein binding and cell uptake; standard across the full ASO length | 过多 PS 增加非特异蛋白结合，是 ASO 类毒性的主要来源之一Excess PS raises non-specific protein binding, a major source of ASO-class toxicity |
| 2′-O-MOE（甲氧乙基methoxyethyl) | 翼部常用，显著提高亲和力与稳定性，毒性相对低Standard in the wings; markedly raises affinity and stability, relatively low toxicity | 放在 gap 里会废掉 RNase H1 切割——gap 必须保持 DNAPlacing it in the gap kills RNase H1 cleavage — the gap must stay DNA |
| 2′-OMe | 翼部可用，成本低，亲和力提升小于 MOEUsable in the wings, cheaper, smaller affinity gain than MOE | 同上，不能进 gapSame as above — never in the gap |
| LNA / cEt | 每颗带来最大的亲和力提升，可缩短 ASO 长度The largest per-residue affinity gain; allows shorter ASOs | 亲和力过强时序列非依赖毒性（如肝毒性）风险上升，新手项目优先 MOE 构型Overly strong affinity raises sequence-independent toxicity (e.g. liver toxicity); beginners should prefer MOE layouts |
ASO gapmer 的标准结构：5′ 翼 3–5 nt（2′ 修饰，提高亲和力与抗酶解）+ 中央 gap 8–10 nt DNA（RNase H1 切割区，不可 2′ 修饰）+ 3′ 翼 3–5 nt（2′ 修饰，同 5′ 翼），硫代磷酸酯（PS）骨架贯穿全长，抗降解 + 助摄取。siRNA 的修饰（一句话版）：体外转染实验用标准未修饰或仅带 dTdT 悬垂的 duplex 即可；免疫敏感细胞可在个别位置加 2′-OMe 降低非特异反应；系统规则见《siRNA 设计指南》第 5 章，不要把 ASO 的全 PS + 翼式修饰搬到 siRNA 上，guide 链重修饰会损害 RISC 装载与切割。递送方式顺带定：脂质体转染 ASO 与 siRNA 通用，起效快，剂量好控制，常规实验首选。裸转染（gymnosis）：PS-ASO 在部分细胞系可不借助试剂、靠长时间高浓度孵育进入细胞，适合转染试剂毒性大的体系，但起效慢（数天）。电转是悬浮细胞和原代细胞的备选，注意电转本身对 lncRNA 表达谱的扰动要设对照。Standard ASO gapmer structure: 5′ wing 3–5 nt (2′-modified, higher affinity and nuclease resistance) + central gap 8–10 nt DNA (the RNase H1 cleavage zone, never 2′-modified) + 3′ wing 3–5 nt (2′-modified, same as the 5′ wing), with a phosphorothioate (PS) backbone throughout for nuclease resistance and uptake. siRNA modification in one line: standard unmodified or dTdT-overhang duplex is enough for in-vitro transfection; immune-sensitive cells can take 2′-OMe at selected positions to dampen non-specific responses; for the full rules see Chapter 5 of the siRNA Design Guide — do not transplant the ASO's full-PS + winged modification onto siRNA, since heavy guide modification harms RISC loading and cleavage. Delivery, decided alongside: lipofection works for ASO and siRNA, is fast and dose-controllable, and is the default for routine work. Gymnosis (naked delivery): in some cell lines PS-ASOs enter cells without reagent by long, high-concentration incubation — good for systems where transfection reagents are toxic, but slow (days). Electroporation is the backup for suspension and primary cells; remember that electroporation itself perturbs the lncRNA expression profile, so include controls for it.
第 7 章 · 验证：qPCR 先过四关，再看功能Chapter 7 · Validation: Pass Four qPCR Gates First, Then Look at Function
lncRNA 的 qPCR 比 mRNA 多几个坑：丰度低、可能没有 poly(A) 尾、反义重叠时分不清链、还容易测到未加工的转录本。四关先过，再谈敲低百分比。1. 逆转录引物选对：多数 lncRNA 有 poly(A) 尾可用 oligo(dT)，但无 poly(A) 的类型（如 eRNA、部分核内 lncRNA）必须用随机六聚体（random hexamers）或基因特异引物；拿不准就用随机引物 + oligo(dT) 混合，或直接全程随机引物。2. 链特异性（反义 / 重叠型必做）：用带标签的链特异引物做逆转录（strand-specific RT），否则测到的是正义基因的信号，敲低数据全盘作废。3. 排除基因组 DNA 与前体：RNA 制备必须 DNase 处理并设 no-RT 对照；引物尽量跨外显子连接处，注意内含子型 lncRNA 和未剪接前体会把"表达量"测虚高。4. 多扩增子互相印证：至少 2 对引物、分别落在不同外显子上，结果一致才算数；单扩增子的"敲低"可能只是该位点被 ASO 占位挡住了引物结合。lncRNA qPCR has extra traps over mRNA: low abundance, possibly no poly(A) tail, strand ambiguity in antisense overlaps, and easy detection of unprocessed transcripts. Pass four gates first, then talk knockdown percentages. 1. Pick the right RT primer: most lncRNAs have a poly(A) tail and take oligo(dT), but poly(A)-negative types (e.g. eRNAs, some nuclear lncRNAs) need random hexamers or gene-specific primers; if unsure, use random primers + oligo(dT) mixed, or random primers throughout. 2. Strand specificity (mandatory for antisense / overlapping types): reverse-transcribe with tagged strand-specific primers (strand-specific RT) — otherwise you measure the sense gene's signal and the knockdown data is worthless. 3. Exclude genomic DNA and precursors: DNase-treat the RNA prep and include no-RT controls; design primers across exon junctions where possible, and remember intronic lncRNAs and unspliced precursors inflate apparent expression. 4. Corroborate with multiple amplicons: at least 2 primer pairs on different exons, and only concordant results count; a single amplicon's "knockdown" may just be the ASO blocking primer binding at that site.
| 对照与定位复核Controls and localization re-checks | 作用Purpose |
|---|
| 阴性对照（同化学打乱 ASO / 非靶向 siRNA）Negative control (same-chemistry scrambled ASO / non-targeting siRNA) | 定义基线，排除化学与转染本身的影响Defines the baseline; excludes chemistry and transfection effects |
| 仅试剂对照（mock）+ 未处理对照Mock (reagent only) + untreated control | 分清递送毒性与靶效应Separates delivery toxicity from target effects |
| 核质分离复核Nuclear / cytoplasmic fractionation re-check | 敲低后分核 / 质测：核内标志（如 U1 snRNA、未剪接 pre-mRNA）与胞质标志（如 GAPDH 成熟 mRNA）先证明分离干净，再看目标在哪个组分被敲下去After knockdown, measure nuclear and cytoplasmic fractions separately: nuclear markers (e.g. U1 snRNA, unspliced pre-mRNA) and cytoplasmic markers (e.g. mature GAPDH mRNA) first prove clean separation, then show which fraction lost the target |
| RNA FISH（有条件时）RNA FISH (when available) | 直观确认定位与敲低，单细胞分辨率能发现"只有部分细胞被敲"的情况Directly confirms localization and knockdown; single-cell resolution reveals cases where only some cells were knocked down |
功能验证：敲低数字不是终点。报告敲低时给出剩余表达量（如"剩余 25%"即敲低 75%），ASO / siRNA 实验在转染后 24–72 h 取样，CRISPRi 看稳定株的持续压制。功能读出回到第 1 章的机制：cis 型先看邻近基因，ceRNA 型看被释放 miRNA 的靶 mRNA，支架 / 向导型看对应复合体的下游输出。最后用第 2 章说的第二种工具复现同一表型，并尽量做 rescue——cis 作用的 lncRNA 常回补不回来，这本身就是机制信息，要如实记录而不是硬凑。Functional validation: the knockdown number is not the finish line. Report knockdown as remaining expression (e.g. "25% remaining" means 75% knockdown); sample ASO / siRNA experiments 24–72 h after transfection, and check sustained repression in CRISPRi stable lines. Choose the functional readout from the Chapter 1 mechanism: cis types — look at neighboring genes first; ceRNA types — look at target mRNAs of the released miRNA; scaffold / guide types — look at the downstream output of the corresponding complex. Finally, reproduce the same phenotype with the second tool from Chapter 2, and attempt a rescue whenever possible — a cis-acting lncRNA often cannot be rescued, which is itself mechanistic information to record honestly rather than force.
第 8 章 · 常见问题排障Chapter 8 · Troubleshooting
| 现象Symptom | 先查Check first | 改法Fix |
|---|
| siRNA 全部不敲低，ASO 能敲No siRNA knocks down; ASO works | 该 lncRNA 是否核内定位（核质分离或文献）Is the lncRNA nuclear (fractionation or literature)? | 不是故障，是工具选错：换 ASO 为主力，siRNA 只用于胞质组分问题Not a fault — wrong tool: switch to ASO as primary; keep siRNA only for the cytoplasmic fraction |
| ASO 也不敲低ASO doesn't knock down either | 阳性对照 ASO 是否工作；靶区是否在重复元件或强结构区；剂量与取样时间（24–72 h 窗）Does the positive-control ASO work? Is the target in a repeat element or strong-structure region? Dose and sampling time (24–72 h window)? | 换到组成型外显子的开放区重新选靶；做剂量梯度；确认递送进了目标细胞Re-target to an open region of a constitutive exon; run a dose gradient; confirm delivery into the target cells |
| 两个扩增子结果对不上Two amplicons disagree | 其中一对引物是否落在试剂靶位点上或单一亚型独有区Does one primer pair sit on the reagent target site or a single-isoform-only region? | 按亚型结构图复核引物位置，统一改到组成型外显子，链特异 RT 重做Re-check primer positions against the isoform map, move both to constitutive exons, redo with strand-specific RT |
| Ct 值很高、重复间乱跳Very high Ct, erratic between replicates | 目标丰度本身低；RNA 起始量；逆转录引物是否匹配（无 poly(A) 却用了 oligo(dT)）Target is inherently low-abundance; RNA input amount; RT primer match (oligo(dT) used on a poly(A)-negative target?) | 加大 RNA 起始量，换随机引物，增加生物学重复后再判读Increase RNA input, switch to random primers, add biological replicates before judging |
| 敲低很好，但没有表型Great knockdown, no phenotype | 功能读出选对没有（cis 看邻居、ceRNA 看靶 mRNA）；敲低持续时间是否覆盖表型窗Is the functional readout right (cis: neighbors; ceRNA: target mRNAs)? Does knockdown duration cover the phenotype window? | 按机制重选读出；延长或重复给药；考虑功能冗余，不要直接判"无功能"Re-pick the readout by mechanism; extend or repeat dosing; consider functional redundancy — do not conclude "no function" outright |
| CRISPR 敲除有表型，ASO 没有CRISPR knockout has a phenotype; ASO doesn't | 删除区段是否含邻居基因元件；敲除株是否有克隆偏差Does the deleted segment contain neighbor-gene elements? Is the knockout clone biased? | 按位点效应处理：多克隆复核 + CRISPRi 对照，表型归因到 DNA 元件而不是 RNATreat as a locus effect: verify across multiple clones + CRISPRi control; attribute the phenotype to the DNA element, not the RNA |
| CRISPRi 压不住CRISPRi won't repress | TSS 定位是否准确；sgRNA 是否落在 -50 至 +300 窗口；dCas9-KRAB 表达量Is the TSS mapped accurately? Do sgRNAs fall in the −50 to +300 window? dCas9-KRAB expression level? | 按 CAGE / 注释重定 TSS，多条 sgRNA 覆盖窗口，先验证 dCas9 体系本身Re-map the TSS from CAGE / annotation, cover the window with multiple sgRNAs, verify the dCas9 system itself first |
| 对照组细胞状态异常Control-group cells look abnormal | mock 与阴性对照是否同样异常；PS-ASO 剂量是否过高Are mock and negative controls equally abnormal? Is the PS-ASO dose too high? | 降剂量、换递送方式；同化学对照出问题说明是化学毒性，不是靶效应Lower the dose, switch delivery; if the same-chemistry control is affected, it is chemical toxicity, not a target effect |
排障顺序口诀：先定位，后体系，再序列。定位与工具匹配、对照正常之后，才有资格说某条序列不好。Troubleshooting order: localization first, then the system, then the sequence. Only once localization matches the tool and controls are clean do you get to call a sequence bad.
第 9 章 · 一个完整设计实例（教学演示）Chapter 9 · A Complete Worked Example (Teaching Demonstration)
先说清：本章是教学演示。目标"LNCX1"与下面所有序列都是按规则构造的教学示例，不对应任何已验证的商业产品，也不能直接下单使用。真实项目的流程与本章完全相同，但序列必须从你的目标转录本现取，并在订购前按当时最新版 RefSeq 重新核对、完成第 5 章双重检查。Up front: this chapter is a teaching demonstration. The target "LNCX1" and every sequence below are teaching examples constructed by the rules; they correspond to no validated commercial product and must not be ordered as-is. A real project follows exactly the same workflow, but sequences must be taken fresh from your target transcript and re-verified against the then-current RefSeq, with the Chapter 5 dual checks completed, before ordering.
任务设定与策略选择：目标是人源细胞系中敲低一个 lncRNA 的全部主要转录本，做 72 h 内表型实验。预实验核质分离显示它约八成在核内 → 按第 2 章决策：ASO gapmer 为主力，siRNA 只作胞质组分的对照试剂，CRISPRi 留作第二工具复现表型。取窗：组成型外显子片段（教学构造，64 nt）：Task and strategy: knock down all major transcripts of an lncRNA in a human cell line for a phenotype experiment within 72 h. Pilot fractionation shows ~80% nuclear localization → per Chapter 2: ASO gapmer as primary, siRNA only as a control reagent for the cytoplasmic fraction, CRISPRi kept as the second tool to reproduce the phenotype. Window: a constitutive-exon fragment (teaching construct, 64 nt):
GACUAUGCAAGCUUACCGAUACGUAGCUAUGCAACGUAUCGAGCUUAACGGCAUUGACUAGCUA
| 候选Candidate | 序列（5′→3′）Sequence (5′→3′) | GC | 评分Score | 决定Decision |
|---|
| ASO-1（gapmer，靶 20 nt）ASO-1 (gapmer, 20 nt target) | 靶Target UGCAACGUAUCGAGCUUAAC
订购序列（DNA）Order sequence (DNA) GTTAAGCTCGATACGTTGCA | 45.0% | — | 下单：5-10-5 MOE gapmer、全 PS 骨架Order: 5-10-5 MOE gapmer, full PS backbone |
| siRNA-1（靶 19 nt）siRNA-1 (19 nt target) | 靶Target CUAUGCAAGCUUACCGAUA
guide UAUCGGUAAGCUUGCAUAG | 42.1% | 8 | 下单（胞质对照）Order (cytoplasmic control) |
| siRNA-2（靶 19 nt）siRNA-2 (19 nt target) | 靶Target AGCUUAACGGCAUUGACUA
guide UAGUCAAUGCCGUUAAGCU | 42.1% | 7 | 下单，与 1 号 seed 不同Order; seed differs from #1 |
| siRNA-3（靶 19 nt）siRNA-3 (19 nt target) | 靶Target AUGCAAGCUUACCGAUACG
guide CGUAUCGGUAAGCUUGCAU | 47.4% | 5 | 降级备用Demoted to backup |
关键判读：siRNA-1 的 guide 链 5′ 端为 U、seed（第 2–8 位 AUCGGUA）GC 42.9%，不对称性成立，得分最高；siRNA-3 的 GC 虽合格，但 guide 5′ 端是 C、seed GC 高达 57.1%，是典型"GC 过关但装载吃亏、脱靶偏高"的序列。ASO-1 的靶段 GC 45.0% 落在 ASO 窗口内，gap 取中央 10 nt DNA，两翼各 5 nt 做 2′-MOE 修饰。下单与验证排程：下单前对 ASO-1、siRNA-1、siRNA-2 做第 5 章全长 BLAST + seed / 近错配双查，确认无重复元件匹配、无对侧基因重叠（本例从略，真实项目不可省）。同批订同化学阴性对照 ASO 与非靶向 siRNA。验证：转染后 48 h 收 RNA，随机引物 + 链特异 RT、2 个扩增子 qPCR；同步做核质分离确认敲低发生在核组分；表型出来后用 CRISPRi（TSS 窗口 4 条 sgRNA）复现。Key calls: siRNA-1's guide starts with U, its seed (positions 2–8, AUCGGUA) is 42.9% GC, and asymmetry holds — the highest score. siRNA-3 passes GC but its guide starts with C and its seed GC is 57.1% — the classic "passes GC but loses on loading and runs hot on off-targets" case. ASO-1's target segment is 45.0% GC, inside the ASO window, with a central 10 nt DNA gap and 5 nt 2′-MOE wings on each side. Ordering and validation schedule: before ordering, run the Chapter 5 full-length BLAST + seed / near-mismatch dual checks on ASO-1, siRNA-1 and siRNA-2, confirming no repeat-element matches and no opposite-strand gene overlap (skipped here for brevity — never skip it in a real project). Order the same-chemistry negative-control ASO and non-targeting siRNA in the same batch. Validation: harvest RNA 48 h after transfection, random primers + strand-specific RT, 2-amplicon qPCR; in parallel, fractionate nucleus and cytoplasm to confirm knockdown in the nuclear fraction; once the phenotype appears, reproduce it with CRISPRi (4 sgRNAs across the TSS window).
附录 · 订购前总清单与参数速查Appendix · Pre-Order Checklist and Quick Parameters
- 目标 lncRNA 在所用细胞中的表达与主要转录本已实测确认，ID + 版本号已记录Target lncRNA expression and major transcripts confirmed by measurement in the cells used; IDs + versions recorded
- 亚细胞定位已知（文献 + 本实验室核质分离 / FISH 至少其一），工具与定位匹配Subcellular localization known (literature + in-house fractionation / FISH, at least one); tool matches localization
- 靶区在组成型外显子上，避开重复元件、常见 SNP 与对侧基因重叠区Target region is on a constitutive exon, avoiding repeat elements, common SNPs, and opposite-strand gene overlap
- 候选共 3 条左右、位置分散，已按第 4 章评分表排序About 3 candidates, positions spread out, ranked on the Chapter 4 scoring table
- siRNA 候选已算 guide 链，5′ 端 A/U、不对称性成立；ASO 候选 gap 8–10 nt、构型与翼修饰写明siRNA candidates' guide strands computed, 5′ end A/U, asymmetry holds; ASO candidates' gap 8–10 nt with layout and wing modifications stated
- 全长 BLAST 已做（短序列模式），无 15 nt 以上连续匹配的非目标转录本Full-length BLAST done (short-sequence mode); no non-target transcript with a ≥ 15 nt contiguous match
- seed（siRNA）/ 近错配谱（ASO）已查，同项目候选 seed 互不相同Seed (siRNA) / near-mismatch spectrum (ASO) checked; candidate seeds in the project all differ
- CRISPRi 项目：TSS 已核准，sgRNA 落在 -50 至 +300 窗口且全基因组脱靶已查CRISPRi projects: TSS verified, sgRNAs in the −50 to +300 window, genome-wide off-targets checked
- 下单规格写明：长度、骨架与修饰（PS / MOE 等）、纯度、量；阴性对照为同化学对照Order spec states length, backbone and modifications (PS / MOE etc.), purity, quantity; negative control is same-chemistry
- 验证方案已定：RT 引物类型与链特异方案、2 个扩增子、核质分离标志物、功能读出Validation planned: RT primer type and strand-specific protocol, 2 amplicons, fractionation markers, functional readout
| 参数Parameter | 常用值Typical value |
|---|
| siRNA 双链区 / GCsiRNA duplex / GC | 19 nt + 3′ dTdT 悬垂overhang / GC 30%–52% |
| ASO gapmer 长度 / 构型ASO gapmer length / layout | 16–20 nt；gap 8–10 nt DNA，翼 3–5 nt 2′ 修饰（如 5-10-5 MOE），全 PS 骨架；GC 40%–60% |
| CRISPRi sgRNA 窗口CRISPRi sgRNA window | TSS -50 至to +300 bp，备 3–5 条筛选 |
| 敲低取样（ASO / siRNA）Knockdown sampling (ASO / siRNA) | 转染后24–72 h 收 RNA，多时间点优先after transfection; multiple time points preferred |
| qPCR 扩增子qPCR amplicons | ≥2 对，不同外显子，结果互证pairs, different exons, mutually corroborating |
| 表型证据标准Phenotype evidence standard | ≥2 种独立试剂（最好跨工具原理）同表型 + 尽量做 rescue≥ 2 independent reagents (preferably across tool types), same phenotype + rescue when possible |
| 中文Chinese | English | 中文Chinese | English |
|---|
| 长链非编码 RNALong non-coding RNA | lncRNA | 反义寡核苷酸Antisense oligonucleotide | ASO |
| 间隙子（缺口体）Gapmer | gapmer | 亚细胞定位Subcellular localization | 亚细胞定位subcellular localization |
| 核糖核酸酶 H1Ribonuclease H1 | RNase H1 | 转录起始位点Transcription start site | TSS |
| CRISPR 干扰CRISPR interference | CRISPRi | CRISPR 敲除CRISPR knockout | CRISPR KO |
| 基因间 lncRNALong intergenic ncRNA | lincRNA | 天然反义转录本Natural antisense transcript | NAT |
| 种子区Seed region | seed region | 脱靶效应Off-target effect | off-target effect |
| 硫代磷酸酯骨架Phosphorothioate backbone | PS backbone | 链特异性逆转录Strand-specific reverse transcription | strand-specific RT |
| 核质分离Nuclear / cytoplasmic fractionation | 核质分离nuclear / cytoplasmic fractionation | 组成型外显子Constitutive exon | constitutive exon |
主要规则来源（可自行查阅原文）：Reynolds et al., Nature Biotechnology 2004 与 Ui-Tei et al., Nucleic Acids Research 2004（siRNA 评分规则，本书沿用）；Bennett & Swayze, Annual Review of Pharmacology and Toxicology 2010（ASO gapmer 与 RNase H 机制）；Gilbert et al., Cell 2013（CRISPRi 原理）与 Larson et al., Nature Protocols 2013（CRISPRi TSS 窗口实践）；Lennox & Behlke, Nucleic Acids Research 2016（ASO 与 siRNA 敲低 lncRNA 的对比）。数据库与工具：GENCODE、RefSeq、Ensembl、lncATLAS、NCBI BLAST、dbSNP。本书为实验方法教育资料，不构成对任何具体产品或厂商的推荐；文中示例序列均为教学构造，不是验证过的商业产品，订购与实验前请按最新数据库（RefSeq / GENCODE）与所在实验室的安全规范复核。本书由 Fern Biotech 实验室指南系列出品。Main rule sources (look up the originals yourself): Reynolds et al., Nature Biotechnology 2004 and Ui-Tei et al., Nucleic Acids Research 2004 (siRNA scoring rules, adopted here); Bennett & Swayze, Annual Review of Pharmacology and Toxicology 2010 (ASO gapmer and RNase H mechanism); Gilbert et al., Cell 2013 (CRISPRi) and Larson et al., Nature Protocols 2013 (CRISPRi TSS-window practice); Lennox & Behlke, Nucleic Acids Research 2016 (ASO vs siRNA for lncRNA knockdown). Databases and tools: GENCODE, RefSeq, Ensembl, lncATLAS, NCBI BLAST, dbSNP. This guide is educational material on experimental method and recommends no specific product or vendor; all example sequences are teaching constructs, not validated commercial products — re-verify against current databases (RefSeq / GENCODE) and your lab's safety rules before ordering or experimenting. From the Fern Biotech Lab Guide Series.
CRISPR sgRNA 设计指南CRISPR sgRNA Design Guide
从靶点选择、序列筛选、脱靶检查到敲除 / 敲入验证——第一次设计 sgRNA 就能走完的完整流程。本书解决的问题：拿到一个基因名之后，不知道选哪段序列、怕脱靶、订了几条都没切动。这本书把设计工作顺序拆成一步步清单，每一步都给出判断标准和检查清单。From target selection, sequence screening, and off-target checking to knockout / knock-in validation — the complete workflow for your first sgRNA design. The problem this guide solves: you have a gene name, but no idea which stretch to pick, a fear of off-targets, and several ordered sgRNAs that never cut. This guide breaks the design work into an ordered checklist, with a decision standard and a verification list at every step.
适用对象：第一次独立设计 sgRNA 的实验人员（细胞实验、基因敲除 / 敲入）Audience: first-time sgRNA designers (cell culture, gene knockout / knock-in)
前置要求：会查 Ensembl / RefSeq，会做基本 PCR 和细胞转染Prerequisites: can use Ensembl / RefSeq; basic PCR and cell transfection
版本 v1.0 · 2026 年 11 月Version 1.0 · November 2026
先记住三个数字：20 nt 是 SpCas9 经典 spacer（向导序列）长度，紧邻 PAM 的那一段才决定切哪里；NGG 是 SpCas9 的 PAM 序列（N 为任意碱基），没有 PAM，序列再匹配也不切；约 3 bp 是双链切口距 PAM 上游的距离，敲入设计时一切距离都从这里量。Three numbers to remember: 20 nt is the classic SpCas9 spacer length — the stretch next to the PAM is what decides where it cuts; NGG is the SpCas9 PAM (N is any base) — without a PAM, no cut no matter how well the sequence matches; and about 3 bp is the distance from the PAM to the double-strand cut, from which every knock-in distance is measured.
最常见的失败原因不是序列，是流程：只订 1 条、没做脱靶检查、验证只看一种方法。本书每一章都在堵这几个口。第 4 章的脱靶检查和最后的订购前清单是硬门槛，没过不要下单。The most common failure is workflow, not sequence: ordering only one sgRNA, skipping off-target checks, validating with a single method. Every chapter plugs one of these holes. The Chapter 4 off-target checks and the final pre-order checklist are hard gates — do not order until you pass them.
第 1 章 · CRISPR/Cas9 原理简介Chapter 1 · How CRISPR/Cas9 Works
1组装：sgRNA 与 Cas9 结合成 RNP 复合物Assembly: sgRNA + Cas9 form the RNP complex
2找 PAM：复合物在基因组上先扫描 PAM 序列Find the PAM: the complex scans the genome for a PAM first
3配对：spacer 与 PAM 旁序列解旋配对（R-loop）Pairing: the spacer unwinds and pairs next to the PAM (R-loop)
4切割：HNH 与 RuvC 结构域各切一条链Cutting: HNH and RuvC each cut one strand
5修复：细胞用 NHEJ 或 HDR 修复双链断裂Repair: the cell fixes the break by NHEJ or HDR
6结果：indel 造成敲除，或按模板实现敲入Outcome: indels knock out, or the template writes a knock-in
CRISPR/Cas9 是把细菌的免疫系统改造成的基因组编辑工具：一条 sgRNA 负责"认地址"，Cas9 蛋白负责"下剪刀"。外源系统不需要从头发明，它直接搭载现成的酶和向导，所以理解原理怎么工作，就理解了后面所有设计规则的来源。PAM 决定能不能切。Cas9 先认 PAM 再验序列。基因组上没有合适的 PAM，这段区域用 SpCas9 就够不着，只能换变体或换区域（第 2、5 章）。seed 区决定认得准不准。spacer 靠 PAM 的约 10–12 nt 是结合的起点，这段错配最容易让切割失败；反过来，脱靶位点只要 seed 区匹配、远端有几个错配，仍可能被切——这是第 4 章脱靶评估要分位置看的原因。结果由修复方式决定，不由 Cas9 决定。切口相同，NHEJ 修复产生随机 indel（插入 / 缺失），多数造成移码而敲除；提供 donor 模板时 HDR 才能按模板敲入。"切得动"和"编得对"是两件事，验证要分开看（第 7 章）。CRISPR/Cas9 is a bacterial immune system repurposed as a genome-editing tool: the sgRNA "reads the address" and the Cas9 protein "does the cutting." Nothing needs inventing from scratch — the system borrows a ready-made enzyme and guide, so understanding how the mechanism works explains every design rule that follows. The PAM decides whether it can cut. Cas9 recognizes the PAM before checking the sequence. If the genome has no suitable PAM in that region, SpCas9 cannot reach it — switch variants or regions (Chapters 2 and 5). The seed decides how precisely it recognizes. The ~10–12 nt of the spacer next to the PAM is where binding starts; mismatches here most easily break cutting — and conversely, an off-target site can still be cut when its seed matches and only the far end mismatches, which is why Chapter 4 grades off-targets by position. The outcome is decided by the repair pathway, not by Cas9. With the same cut, NHEJ repair produces random indels (insertions / deletions), most of which frameshift and knock out; only with a donor template can HDR write a knock-in. "It cuts" and "it edits correctly" are two different things — validate them separately (Chapter 7).
| 四种常见用法Four common uses | 需要什么What it needs | 结果Result |
|---|
| 敲除 knockoutKnockout | sgRNA + Cas9 | NHEJ 产生 indel，移码失活NHEJ produces indels; frameshift inactivates the gene |
| 敲入 knock-inKnock-in | 再加 donor 模板Plus a donor template | 点突变、标签或片段按模板写入Point mutation, tag, or fragment written from the template |
| CRISPRi | dCas9（无切割活性）+ 效应结构域dCas9 (no cutting activity) + effector domain | 不切 DNA，只压低转录No DNA cutting; transcription repressed |
| CRISPRa | dCas9 + 转录激活结构域dCas9 + transcriptional activation domain | 不切 DNA，抬高内源基因表达No DNA cutting; endogenous gene expression raised |
本书范围以 SpCas9 介导的敲除为主线，敲入在第 2、7 章单独标注差异；CRISPRi/a 只在第 5 章对照提及，设计规则不完全通用。This guide centers on SpCas9-mediated knockout; knock-in differences are flagged separately in Chapters 2 and 7. CRISPRi/a appear only as a comparison in Chapter 5 — their design rules are not fully interchangeable.
术语约定：向导序列称 spacer（间隔序列），与基因组互补配对的那 20 nt；PAM 指 protospacer adjacent motif（原间隔序列邻近基序），在基因组上、不在 sgRNA 里；seed 区（种子区）指 spacer 中紧邻 PAM 的约 10–12 nt，错配最敏感。文中序列除特别注明外，都按 5′→3′ 书写。Terminology: the guide sequence is the spacer — the 20 nt that pairs with the genome. PAM (protospacer adjacent motif) sits on the genome, not in the sgRNA. The seed region is the ~10–12 nt of the spacer next to the PAM, where mismatches matter most. Sequences are written 5′→3′ unless noted.
第 2 章 · 靶点选择：区域与 PAMChapter 2 · Target Selection: Region and PAM
设计第一步不是挑序列，是挑区域。区域选错，后面 GC、评分都合格也救不回来：要么所有转录本共有的外显子没覆盖，要么切口离想改的位置太远。PAM 怎么认（以 SpCas9 为例）：在目标区域双链上找 5′-NGG-3′；紧邻 PAM 上游的 20 nt 就是 spacer。反链上的 CCN 同样可用（反向读即为 NGG）。切口在 PAM 上游约 3 bp 处，两条链都从这里量距离。spacer 按 5′→3′、与 PAM 同向书写；订购合成 sgRNA 时给的就是这条序列，PAM 不写进去。The first design step is not picking a sequence — it is picking a region. Pick the wrong region and no amount of passing GC and scores saves it: either the exon shared by all transcripts is missed, or the cut lands too far from the intended edit. How to read a PAM (SpCas9): find 5′-NGG-3′ on either strand of the target region; the 20 nt immediately upstream of the PAM is the spacer. CCN on the opposite strand works the same (read in reverse it is NGG). The cut sits ~3 bp upstream of the PAM; measure both strands from there. Write the spacer 5′→3′ in the same direction as the PAM; when ordering synthetic sgRNA, this is the sequence you give — the PAM is not included.
- 敲除：优先选所有主要转录本共有的早期外显子（constitutive exon，共同外显子），一套 sgRNA 覆盖全部亚型Knockout: prefer an early exon shared by all major transcripts (constitutive exon) so one sgRNA set covers all isoforms
- 敲除：位置放在 CDS（编码区）前半段为宜；太靠后（近 C 端）的移码可能只截掉一小段，蛋白还剩部分功能Knockout: aim at the first half of the CDS; a frameshift near the C-terminus may only truncate a short piece and leave partial function
- 敲除：避开可变剪接外显子、重复序列和低复杂度区；查 dbSNP，避开人群常见 SNP 位点——你的细胞系若正好带这个 SNP，配对直接受损Knockout: avoid alternatively spliced exons, repeats, and low-complexity regions; check dbSNP and avoid common population SNPs — if your cell line carries that SNP, pairing is directly damaged
- 敲除：每个基因选 3 条，分布在 2 个以上外显子上，不要 3 条挤在同一小段Knockout: pick 3 sgRNAs per gene, spread across 2+ exons — never cluster all three in one small stretch
- 敲入：切口离编辑位点的距离决定成败——切口距要改的碱基最好在 10 bp 以内，越近 HDR 效率越高；超过约 20–30 bp 先换 sgRNA，不要靠加大 donor 用量硬冲Knock-in: cut-to-edit distance decides success — the cut should sit within 10 bp of the base to change (closer means higher HDR efficiency); beyond ~20–30 bp, switch sgRNAs instead of forcing it with more donor
- 敲入：donor 的同源臂以切口为中心向两侧延伸设计（ssODN 常用每侧约 30–60 nt）Knock-in: design donor homology arms extending both ways from the cut (ssODN typically ~30–60 nt per side)
另外注意：敲入成功后若 PAM 或 seed 区原样保留，Cas9 会把已编辑好的等位基因再切一遍。donor 里顺手把 PAM 或 seed 区改成同义突变（不改变氨基酸），是标准做法。One more thing: after a successful knock-in, if the PAM or seed region is left unchanged, Cas9 will re-cut the edited allele. The standard fix is to build a synonymous mutation (no amino-acid change) into the PAM or seed region in the donor.
第 3 章 · 设计规则：序列与评分Chapter 3 · Design Rules: Sequences and Scoring
区域定下后，在窗口内滑动取每一条可用的 spacer，先过硬规则，再用评分排序。顺序反了，比如先看中一条高分序列再发现它带常见 SNP，基本都会翻车。Once the region is set, slide through the window to collect every usable spacer, pass them through the hard rules first, then rank by score. Reverse the order — fall for a high-scoring sequence first and discover its common SNP later — and the project usually fails.
| 硬规则（先淘汰，不讲情面）Hard rules (eliminate first, no exceptions) | 标准与原因Standard and reason |
|---|
| GC 含量 40%–60%GC content 40%–60% | 太低结合弱、活性差；太高容易形成二级结构且脱靶风险上升Too low: weak binding, poor activity. Too high: secondary structure and higher off-target risk |
| 含 TTTT 直接淘汰（用 U6 启动子表达时）TTTT: eliminate outright (U6 expression) | 连续 4 个 T 是 Pol III 的终止信号，sgRNA 会被截短。化学合成 sgRNA 不受此限，但避开连续 4 个以上相同碱基仍是稳妥做法Four consecutive Ts terminate Pol III transcription, truncating the sgRNA. Chemically synthesized sgRNA is exempt, but avoiding 4+ identical bases in a row is still the safe practice |
| 5′ 端偏好 G（U6 表达时）5′ end prefers G (U6 expression) | U6 启动子转录起始偏好 G；候选 5′ 不是 G 时可在前端加一个 G（会多出 1 nt，多数情况下可接受）U6 transcription starts preferentially at G; if a candidate doesn't start with G, add one at the front (adds 1 nt, usually acceptable) |
| seed 区 GC 别过高Don't let seed GC run high | PAM 近端过稳会提高对错配的容忍，脱靶更难压An over-stable PAM-proximal end tolerates mismatches more, making off-targets harder to suppress |
| 避开已知常见 SNP、微卫星重复和与基因组其他位置高度同源的区段Avoid known common SNPs, microsatellite repeats, and segments highly homologous to elsewhere in the genome | 同源区段是脱靶的直接来源Homologous segments are direct off-target sources |
| 评分工具：用来排序，不用来打包票Scoring tools: for ranking, not for guarantees | 给什么分What it scores | 备注Notes |
|---|
| Rule Set 2 / Azimuth | on-target 活性预测On-target activity prediction | 主流设计工具的底层模型之一One of the underlying models of mainstream design tools |
| CRISPOR | 活性 + 脱靶位点一览Activity + off-target site overview | 同时给 CFD / MIT 脱靶分，查脱靶顺手Gives CFD / MIT off-target scores too — handy for off-target checks |
| CHOPCHOP | 活性排序 + 可视化Activity ranking + visualization | 多物种、多核酸酶，选区时方便Multi-species, multi-nuclease; convenient when picking regions |
| Cas-OFFinder | 只查脱靶Off-targets only | 可设错配数，还能查 bulge 型脱靶Mismatch count adjustable; also finds bulge-type off-targets |
| 厂商在线设计器Vendor online designers | 活性 + 现货库Activity + in-stock libraries | 有验证过的现货序列时优先考虑Prefer when validated in-stock sequences exist |
分数是同一基因内候选之间的相对排序，不是跨基因可比的绝对活性保证。同一条序列换细胞类型、换递送方式，实际效率可以差几倍。正确用法：硬规则筛完 → 按分数排前 3 → 3 条一起订、一起测，最后用实验结果定主选。Scores are relative rankings among candidates for the same gene, not absolute activity guarantees comparable across genes. The same sequence in a different cell type or delivery method can differ in efficiency several-fold. Correct use: filter by hard rules → take the top 3 by score → order and test all three together, and let the experiment pick the winner.
第 4 章 · 脱靶检查：全基因组评估Chapter 4 · Off-Target Checking: Genome-Wide Assessment
Cas9 容忍错配：PAM 远端有 1–3 个错配时仍可能切割，个别位点带插入 / 缺失型错配（bulge）也能切。脱靶检查不是下单前点一下的仪式，而是决定这条 sgRNA 能不能用的筛选步骤。检查流程（5 步）：1. 拿最终要订的 spacer（含加 G 等任何改动后的版本）做查询，改动前的序列查了等于没查。2. 用 CRISPOR、CHOPCHOP 或 Cas-OFFinder 对与你实验物种一致的最新版参考基因组做全基因组搜索，允许至少 3 个、稳妥看 4 个错配。3. 按位置给脱靶位点分级：在其他基因外显子里的最危险，其次内含子，基因间区相对低。4. 用 CFD 或 MIT 分给每个脱靶位点打分排序；高分（高风险）位点逐个看。5. 实验做完后，对排前的高风险位点做靶向扩增子测序确认——预测没查到不等于实验没发生。Cas9 tolerates mismatches: it can still cut with 1–3 mismatches at the PAM-distal end, and a few sites with insertion / deletion-type mismatches (bulges) can be cut too. Off-target checking is not a pre-order ritual — it is the screen that decides whether an sgRNA is usable. The 5-step check: 1. Query with the final sequence to be ordered (including any added G or other change) — checking the pre-change sequence counts for nothing. 2. Search the whole genome of the current reference build for your species with CRISPOR, CHOPCHOP, or Cas-OFFinder, allowing at least 3 mismatches (4 to be safe). 3. Grade off-target sites by position: most dangerous in another gene's exon, then intron, then intergenic. 4. Score and rank every off-target site with CFD or MIT scores; inspect high-score (high-risk) sites one by one. 5. After the experiment, confirm the top high-risk sites by targeted amplicon sequencing — not predicted does not mean not cut.
| 情形Situation | 风险Risk | 处置Action |
|---|
| seed 区错配（≥1 个）Mismatch in the seed (≥ 1) | 低Low | 一般可接受，记录在案Usually acceptable; record it |
| 仅 PAM 远端 1–2 个错配Only 1–2 PAM-distal mismatches | 中Medium | 看所在位置：落在外显子则换序列Depends on position: switch sequences if it lands in an exon |
| 全长 0–1 个错配的第二位点A second site with 0–1 mismatches genome-wide | 高High | 直接换 sgRNA，不要赌Switch sgRNAs outright; don't gamble |
| 脱靶位点在同家族基因上Off-target site on a same-family gene | 高High | 换序列；验证时也要单独查该基因Switch sequences; check that gene separately during validation too |
预测工具按参考基因组算账：你的细胞系自己的 SNP 可能造出参考基因组里没有的新脱靶位点。要求高的实验（比如要建稳定株、后续要发表），重要脱靶位点请按细胞系背景再复核一遍，不要只信网页工具的默认结果。Prediction tools calculate on the reference genome: your own cell line's SNPs can create new off-target sites that don't exist in the reference. For demanding work (stable lines, results headed for publication), re-check important off-target sites against the cell line's background — don't trust a web tool's defaults alone.
压脱靶的四个手段（设计阶段就能用）：选本身脱靶谱干净的序列、用高保真 Cas9（第 5 章）、用 RNP 缩短 Cas9 在细胞里的暴露时间（第 6 章）、把剂量降到够用的最低值。双切口 nickase 方案把特异性再抬一档，代价是要设计一对 sgRNA。Four ways to push off-targets down (usable at the design stage): pick a sequence with a clean off-target profile, use a high-fidelity Cas9 (Chapter 5), use RNP to shorten Cas9 exposure in the cell (Chapter 6), and drop the dose to the lowest that works. A double-nickase design raises specificity another notch, at the cost of designing a sgRNA pair.
第 5 章 · Cas9 变体与高保真酶选择Chapter 5 · Cas9 Variants and High-Fidelity Enzymes
野生型 SpCas9 不是唯一选项。选酶只回答两个问题：目标位置有没有它认的 PAM，以及你的实验对脱靶的容忍度有多低。Wild-type SpCas9 is not the only option. Choosing an enzyme answers two questions: does the target position have a PAM it recognizes, and how little off-targeting can your experiment tolerate?
| 酶Enzyme | PAM | 特点Features | 什么时候选When to choose it |
|---|
| SpCas9（野生型wild-type) | NGG | 活性基准，工具最全The activity benchmark; most tools available | 常规敲除的默认选项Default for routine knockout |
| eSpCas9(1.1) | NGG | 高保真，脱靶明显下降High fidelity, off-targets clearly reduced | 脱靶敏感、活性要求不极端Off-target-sensitive work where activity demands are not extreme |
| SpCas9-HF1 | NGG | 高保真，个别位点活性降High fidelity; activity drops at some sites | 同上，下单前多备一条 sgRNASame as above; keep an extra sgRNA ready before ordering |
| HypaCas9 | NGG | 高保真，seed 检验更严High fidelity; stricter seed checking | 脱靶位点 seed 区匹配时优先Prefer when off-target sites match in the seed |
| HiFi Cas9 | NGG | 单点突变高保真，有 RNP 产品Single-point-mutation high fidelity; RNP products available | 做 RNP 递送又要压脱靶RNP delivery with off-targets to suppress |
| Cas9 切口酶Cas9 nickase | NGG ×2 | 只切单链，需成对 sgRNACuts one strand only; needs a sgRNA pair | 特异性优先、能找到成对位点Specificity first, and a paired site is available |
| SaCas9 | NNGRRT | 蛋白更小，可装 AAVSmaller protein, fits AAV | 体内递送、AAV 载量受限In-vivo delivery, AAV capacity limited |
| Cas12a | TTTV | 切口为黏性末端，AT 富集区可用Sticky-end cuts; works in AT-rich regions | 目标区缺 NGG、富含 AT 时Target region lacks NGG and is AT-rich |
| SpG / SpRY | NGN / 近无限制NGN / near-unrestricted | PAM 兼容范围大Broad PAM compatibility | 敲入切口位置被 PAM 卡死时When the knock-in cut position is blocked by PAM availability |
| dCas9 | 同来源酶Same as source enzyme | 无切割活性，只结合No cutting activity; binds only | CRISPRi / CRISPRa，不做敲除not knockout |
选择建议：普通细胞系敲除：野生型 + RNP 就够了，先把流程跑通，不要一上来堆高保真酶排查变量。原代细胞、干细胞、后续要做动物或发表级结论：高保真酶 + RNP + 第 4 章的实验验证，三件一起上。高保真酶的代价是部分 sgRNA 活性下降：换酶等于换了系统，原来测过的效率不能直接沿用，要重新测。PAM 卡死敲入位点时，顺序是：先在正负链扩大找 NGG → 再考虑 SpG → 还不行才换 Cas12a 家族。Selection advice: routine knockout in a standard cell line: wild-type + RNP is enough — get the workflow running first instead of stacking high-fidelity enzymes and multiplying variables. Primary cells, stem cells, or work headed for animals or publication: high-fidelity enzyme + RNP + the Chapter 4 experimental validation, all three together. The price of a high-fidelity enzyme is lower activity for some sgRNAs: switching enzymes means switching systems, so previously measured efficiencies don't carry over — re-measure. When a PAM blocks the knock-in cut site, the order is: widen the NGG search to both strands → then consider SpG → only then switch to the Cas12a family.
第 6 章 · 递送方式：把 RNP 送进细胞Chapter 6 · Delivery: Getting RNP into the Cell
同一条 sgRNA，递送方式不同，效率和脱靶可以差出一个量级。递送选错，第 3 章的分数再高也测不出来。The same sgRNA delivered differently can differ in efficiency and off-targets by an order of magnitude. Pick the wrong delivery and no Chapter 3 score will save the measurement.
| 方式Method | 起效与持续Onset and duration | 脱靶 / 风险Off-targets / risks | 适用Best for |
|---|
| RNP | 立即起效，约 1–2 天降解Immediate onset, degraded in ~1–2 days | 暴露时间短，脱靶低；无整合风险Short exposure, low off-targets; no integration risk | 多数细胞系与原代细胞的默认首选Default first choice for most cell lines and primary cells |
| 质粒Plasmid | 起效慢，表达持续数天Slow onset, expression lasts days | 暴露长，脱靶偏高；有随机整合可能Long exposure, higher off-targets; possible random integration | 好转染的细胞系、需带筛选标记时Easy-to-transfect lines, or when a selection marker is needed |
| mRNA | 起效快，持续约 1 天Fast onset, lasts ~1 day | 介于 RNP 与质粒之间Between RNP and plasmid | 难做 RNP 电转、又忌质粒的场景When RNP electroporation is impractical and plasmid is undesirable |
| 慢病毒Lentivirus | 整合后长期表达Long-term expression after integration | 持续表达脱靶累积；整合位点风险Off-targets accumulate with continuous expression; integration-site risk | 难转染细胞、文库筛选（pooled screen）Hard-to-transfect cells; pooled library screens |
| AAV | 体内递送主力Main in-vivo delivery vehicle | 载量小，常需双载体或 SaCas9Small cargo; often needs dual vectors or SaCas9 | 体内实验；体外细胞实验一般不用In-vivo work; rarely used for in-vitro cells |
递送要点：RNP 组装：sgRNA 与 Cas9 按厂商推荐摩尔比（常见约 1.2:1 到 3:1）室温孵育成复合物再电转，具体比例以所用产品说明书为准。敲入时 donor 与 RNP 同时递送；ssODN 适合点突变和小标签，长片段敲入用质粒或 AAV donor，设计见第 2 章。电转程序、细胞密度和恢复时间按细胞类型单独优化；第一次做先设一条已知高活性的阳性对照 sgRNA，把递送问题和序列问题分开。难转染细胞不要硬加剂量：先优化递送条件（电转参数、脂质体比例），再回头怀疑序列。做 pooled 文库筛选是另一套流程（慢病毒、低 MOI 感染），本书的单条设计规则只覆盖其中的序列部分。Delivery points: RNP assembly: incubate sgRNA and Cas9 at the vendor-recommended molar ratio (commonly ~1.2:1 to 3:1) at room temperature to form the complex before electroporation — follow the specific product's instructions. For knock-in, deliver donor and RNP together; ssODN suits point mutations and small tags, while long-fragment knock-in uses plasmid or AAV donors (design in Chapter 2). Optimize electroporation program, cell density, and recovery time per cell type; on the first run, include a known high-activity positive-control sgRNA to separate delivery problems from sequence problems. For hard-to-transfect cells, don't just push the dose — optimize delivery conditions (electroporation parameters, lipid ratios) before suspecting the sequence. Pooled library screening is a different workflow (lentivirus, low-MOI infection); this guide's single-sgRNA design rules cover only its sequence part.
为什么 RNP 是默认首选：Cas9 在细胞里待得越久，切对的机会多一点，切错的机会也多一点。RNP 干完活就降解，等于同时买了效率和特异性；质粒持续表达几天，脱靶账是按天累积的。Why RNP is the default first choice: the longer Cas9 stays in the cell, the more chances it gets to cut right — and to cut wrong. RNP degrades after doing its job, buying efficiency and specificity together; a plasmid expressing for days accumulates off-targets day by day.
第 7 章 · 验证：敲除与敲入怎么确认Chapter 7 · Validation: Confirming Knockout and Knock-In
编辑完成的信号要分三层看：DNA 层（切没切、indel 长什么样）、蛋白层（目标蛋白还在不在）、功能层（表型对不对）。只看一层就下结论，是验证环节最常见的错误。标准流程（敲除，群体细胞）：1. 递送后 48–72 h 收细胞，提基因组 DNA（留一份未编辑对照）。2. PCR 扩增靶位点：引物放在切口两侧，产物约 400–800 bp，且引物不能落在 sgRNA 结合区内。3. T7E1（或 Surveyor）初筛：有切割条带说明群体里有 indel；它只能半定量。4. Sanger 测序 + TIDE / ICE 分解：得到 indel 总比例和主要 indel 类型的构成。5. 分单克隆：逐个克隆测序定基因型（纯合 / 杂合 / 哪种 indel），再做蛋白层确认。Read the editing signal in three layers: DNA (was it cut, and what do the indels look like), protein (is the target protein still there), and function (is the phenotype right?). Concluding from one layer alone is the most common validation mistake. Standard knockout workflow (cell population): 1. Harvest cells 48–72 h after delivery and extract genomic DNA (keep an unedited control). 2. PCR-amplify the target site: primers on both sides of the cut, product ~400–800 bp, and primers must not sit inside the sgRNA binding region. 3. T7E1 (or Surveyor) prescreen: cut bands mean indels are present in the population; this is semi-quantitative only. 4. Sanger sequencing + TIDE / ICE decomposition: total indel percentage and the makeup of the main indel types. 5. Isolate single clones: sequence each clone to fix the genotype (homozygous / heterozygous / which indel), then confirm at the protein layer.
| 方法Method | 测什么What it measures | 灵敏度 / 局限Sensitivity / limits | 定位Role |
|---|
| T7E1 | 错配存在与否Presence of mismatches | 检测限约 1%–5%；纯合相同 indel 会漏检Detection limit ~1%–5%; homozygous identical indels are missed | 快速初筛Fast prescreen |
| Sanger + TIDE/ICESanger + TIDE / ICE | indel 比例与谱型Indel percentage and profile | 低比例（约低于 5%）分解不可靠Unreliable for low proportions (roughly under 5%) | 群体定量Population quantification |
| 扩增子 NGSAmplicon NGS | 每个等位基因的精确序列Exact sequence of every allele | 成本高，需生信分析Costly; needs bioinformatics | 发表级定量、脱靶验证Publication-grade quantification and off-target validation |
| Western blotWestern blot | 目标蛋白是否消失Whether the target protein is gone | 依赖好抗体；蛋白半衰期长时要等Needs a good antibody; long-half-life proteins take time | 敲除的蛋白层证据Protein-level evidence of knockout |
| qPCR | mRNA 水平mRNA level | 测的是转录本，不是编辑本身Measures the transcript, not the edit itself | 辅助证据，不能单独用Supporting evidence only |
三个最常见的误判：（1）T7E1 阴性就说没编辑——纯合 indel 和低比例编辑都会被它漏掉；（2）qPCR 下降就说敲除了——移码常触发无义介导的 mRNA 降解，mRNA 降不等于蛋白没了，反过来 mRNA 不变也不等于没敲；（3）群体测序有 indel 就说敲除株建好了——没有单克隆基因型和蛋白证据，株还没建成。The three most common misjudgments: (1) "T7E1 negative means no editing" — homozygous indels and low-frequency edits both escape it. (2) "qPCR went down, so it's knocked out" — frameshifts often trigger nonsense-mediated mRNA decay, so mRNA down does not mean protein gone, and unchanged mRNA does not mean no knockout. (3) "The population has indels, so the knockout line is built" — without single-clone genotypes and protein evidence, the line is not built.
敲入多两步：接头 PCR（junction PCR）：一条引物在同源臂外侧基因组上、一条在插入片段内，证明插入发生在正确位置，而不是随机整合。Sanger 跨臂测序：读穿整个编辑区，逐碱基确认目标突变、PAM 同义突变都在，且没有附带 indel。Knock-in needs two extra steps: junction PCR — one primer on genomic DNA outside the homology arm, one inside the insert — proving the insertion landed in the right place, not a random integration. Sanger sequencing across the arms: read through the entire edited region to confirm base by base that the intended mutation and the PAM synonymous mutation are present, with no accompanying indels.
第 8 章 · 常见问题排障Chapter 8 · Troubleshooting
排障顺序永远是：先分清问题出在递送、序列还是验证，再动手。阳性对照 sgRNA 切得动而你的切不动，是序列问题；连对照都切不动，是递送问题，不要回去改序列。Troubleshooting order is fixed: first tell whether the problem is delivery, sequence, or validation — then act. If the positive-control sgRNA cuts but yours doesn't, it is a sequence problem; if even the control doesn't cut, it is a delivery problem — don't go back and redesign the sequence.
| 现象Symptom | 常见原因Common causes | 先查什么Check first |
|---|
| 效率为零（TIDE 无信号）Zero efficiency (no TIDE signal) | 递送失败；sgRNA 无活性；细胞系靶位点有 SNPDelivery failed; inactive sgRNA; SNP at the target site in this cell line | 阳性对照结果；测序确认靶位点序列Positive-control result; sequence the target site to confirm |
| T7E1 阴性、测序却有 indelT7E1 negative but indels by sequencing | 纯合相同 indel 或低比例编辑被漏检Homozygous identical indels or low-frequency edits missed | 属正常现象，以测序为准（第 7 章）Normal phenomenon; trust the sequencing (Chapter 7) |
| indel 高但蛋白还在High indels but protein remains | indel 非移码（3 的倍数）；蛋白半衰期长Indels are in-frame (multiples of 3); long protein half-life | 看 indel 谱型；延长培养后再测 WesternRead the indel profile; extend culture and re-test by Western |
| 有蛋白消失但无表型Protein gone but no phenotype | 旁系同源基因代偿；表型检测不对路Paralog compensation; wrong phenotypic readout | 查同家族基因表达；换功能读出Check same-family gene expression; switch functional readout |
| 单克隆全是杂合All clones heterozygous | 细胞多倍体；编辑效率只够切一条等位基因Polyploid cells; efficiency only enough to cut one allele | 对纯合克隆再做一轮递送；多挑克隆Re-deliver into a heterozygous clone; pick more clones |
| 预测的高风险脱靶位点被切Predicted high-risk off-target site cut | 序列本身脱靶谱差；剂量过高、暴露过长Poor off-target profile of the sequence; dose too high, exposure too long | 换 sgRNA 或高保真酶；改 RNP 并降剂量Switch sgRNA or high-fidelity enzyme; switch to RNP and lower the dose |
| 递送后细胞大量死亡Massive cell death after delivery | 电转条件过强；donor 毒性；DSB 负荷过高Electroporation too harsh; donor toxicity; too many double-strand breaks | 降电转强度与 DNA 用量；分开优化 donorLower electroporation intensity and DNA dose; optimize the donor separately |
| 敲入效率始终很低Knock-in efficiency stays very low | 切口离编辑位点太远；donor 同源臂设计不当Cut too far from the edit site; donor homology arms poorly designed | 回第 2 章换切口更近的 sgRNAGo back to Chapter 2 and pick an sgRNA with a closer cut |
| 换个细胞系效率骤降Efficiency collapses in a new cell line | 染色质开放程度不同；SNP；转染条件不同Different chromatin openness; SNPs; different transfection conditions | 先测靶位点序列，再谈换序列Sequence the target site first, then talk about switching sequences |
排障时一次只改一个变量，并把每轮的 sgRNA 批次、递送条件和验证结果记在同一张表上。没有记录的排障，下一轮还会踩同一个坑。Change one variable at a time during troubleshooting, and record each round's sgRNA batch, delivery conditions, and validation results in the same table. Troubleshooting without records repeats the same mistake next round.
第 9 章 · 一个完整的设计实例（教学演示）Chapter 9 · A Complete Worked Example (Teaching Demonstration)
先说清：本章是教学演示。演示序列是按规则构造的教学示例，用来走通每一步计算，不是验证过的产品序列，也不对应任何真实基因的已验证 sgRNA。真实项目中，第 1 步从最新版 Ensembl / RefSeq 取你的目标序列之后，其余步骤与本章完全相同；订购前必须按最新基因组版本重新核对并完成第 4 章检查。Up front: this chapter is a teaching demonstration. The sequences here are teaching examples constructed by the rules, to walk through every calculation — they are not validated product sequences and correspond to no validated sgRNA for any real gene. In a real project, once step 1 takes your target sequence from the current Ensembl / RefSeq, every later step is identical to this chapter; before ordering you must re-verify against the current genome build and complete the Chapter 4 checks.
任务设定：目标是在人源细胞系中敲除某个基因的全部主要转录本，用于后续功能实验。方案已定：SpCas9 野生型、RNP 电转递送、递送后 72 h 收样验证。预算够订 3 条候选 + 阳性对照。步骤 1–2：定转录本，选区域。查 Ensembl：记录目标基因主转录本编号与版本号，下载 CDS 序列，确认几个主要转录本共有的早期外显子区段。查 dbSNP：标出共有外显子区内的常见 SNP 位点，选区时整段绕开。在共有外显子的前、中、后三段各取候选，保证 3 条候选分布分散（第 2 章清单逐项过）。步骤 3：滑窗 + 硬规则初筛。在选定区段双链上找 NGG / CCN，在每个 PAM 上游取 20 nt spacer，先淘汰 GC 不在 40%–60%、含连续 4 个相同碱基、落在常见 SNP 上的序列。本例进入评分阶段的 3 条如下（spacer 方向，5′→3′）：Task: knock out all major transcripts of a gene in a human cell line for downstream functional work. Plan fixed: wild-type SpCas9, RNP electroporation, harvest and validate 72 h after delivery. Budget covers 3 candidates + a positive control. Steps 1–2: fix the transcript, pick the region. Check Ensembl: record the main transcript ID and version, download the CDS, and confirm the early exon segment shared by the major transcripts. Check dbSNP: mark common SNP sites inside the shared exons and route around them entirely. Take candidates from the front, middle, and back of the shared exons so the 3 candidates are spread out (Chapter 2 checklist, item by item). Step 3: sliding window + hard-rule screen. Find NGG / CCN on both strands of the chosen segment, take the 20 nt spacer upstream of each PAM, and first eliminate anything with GC outside 40%–60%, four identical bases in a row, or a common SNP. The 3 sequences reaching scoring in this example (spacer orientation, 5′→3′):
| 候选Candidate | spacer 序列（20 nt）Spacer sequence (20 nt) | PAM | GC |
|---|
| 候选 1Candidate 1 | GAGTCCGAGCAGAAGAAGAA | AGG | 50.0% |
| 候选 2Candidate 2 | CTTGACCTGGAGGAGAGCAG | TGG | 60.0% |
| 候选 3Candidate 3 | GACATCAACCACACCTACCT | 反链 CCNCCN on the opposite strand | 50.0% |
GC 计算（候选 1）：序列含 G 7 个、C 3 个，共 10 个，10 / 20 = 50.0%，落在窗口正中。三条都过了硬规则——下一步的评分和脱靶才决定下单顺序。GC calculation (candidate 1): the sequence has 7 Gs and 3 Cs, 10 total, 10 / 20 = 50.0% — dead center of the window. All three passed the hard rules; scoring and off-target checks now decide the order.
| 步骤 4：评分排序Step 4: score and rank | 活性预测分*Predicted activity* | 加分 / 扣分点Strengths / deductions | 决定Decision |
|---|
| 1 | 高High | GC 居中，seed 区无同聚序列，位置在共有外显子前段GC centered, no homopolymer in the seed, located in the front of the shared exon | 首选下单First choice to order |
| 3 | 中Medium | GC 合格，反链靶点，与候选 1 的 seed 完全不同GC passes; opposite-strand target; seed completely different from candidate 1 | 独立第二条，下单Independent second — order |
| 2 | 中Medium | GC 顶格 60%，seed 区 GC 偏高，脱靶余量小GC at the 60% ceiling, seed GC on the high side, little off-target margin | 降级备用Demoted to backup |
* 分档为教学演示的相对排序示意，不是任何工具的真实输出数值；真实项目请以你所用工具对你的真实序列给出的分数为准。步骤 5：脱靶检查（第 4 章流程）。对候选 1、3 用全基因组工具按 ≤4 错配搜索：候选 1 的近匹配位点 seed 区均有错配，且不在外显子内，通过。候选 3 有一个 2 错配位点落在基因间区，记录在案，实验后用扩增子测序复查该位点。候选 2 备用：若前两条活性不够再启用，启用前同样要先过完本步。步骤 6–7：下单与验证计划。下单规格：化学合成 sgRNA（两端带保护性修饰的版本更耐降解）× 候选 1、3，外加一条已验证的阳性对照 sgRNA；Cas9 蛋白与电转试剂按 RNP 方案备齐。到货后验证顺序：RNP 电转 → 72 h 提 DNA → T7E1 初筛 → Sanger + TIDE 定量 → 分单克隆测序定基因型 → Western 确认蛋白消失（第 7 章）。判定标准写在实验前：群体 indel 比例、单克隆纯合移码、Western 无条带，三项都到才算敲除完成。本例的要点不在这几条序列，而在顺序：选区 → 硬规则 → 评分排序 → 脱靶双查 → 才下单。顺序反了，比如先看中一条序列再倒推理由，基本都会翻车。* These grades are relative rankings for teaching demonstration only, not real output values from any tool; in a real project, use the scores your tool gives your real sequences. Step 5: off-target checks (Chapter 4 workflow). Search genome-wide for candidates 1 and 3 at ≤ 4 mismatches: candidate 1's near-match sites all carry seed mismatches and sit outside exons — pass. Candidate 3 has one 2-mismatch site in an intergenic region; record it and re-check it by amplicon sequencing after the experiment. Candidate 2 stays as backup: if the first two underperform, activate it only after it passes this same step. Steps 6–7: ordering and validation plan. Order spec: chemically synthesized sgRNA (end-protected versions resist degradation better) × candidates 1 and 3, plus one validated positive-control sgRNA; Cas9 protein and electroporation reagents per the RNP protocol. Validation after delivery: RNP electroporation → DNA extraction at 72 h → T7E1 prescreen → Sanger + TIDE quantification → single-clone sequencing for genotype → Western to confirm protein loss (Chapter 7). Write the pass criteria before the experiment: population indel percentage, homozygous frameshift in single clones, and no Western band — all three must hold for the knockout to count as complete. The lesson of this example is not the sequences but the order: region → hard rules → score and rank → dual off-target checks → only then order. Reverse the order — pick a favorite sequence first and reason backwards — and it usually fails.
附录 · 订购前清单、参数速查与术语对照Appendix · Pre-Order Checklist, Quick Parameters, and Glossary
- 目标转录本 ID 与版本号已记录，靶区与"敲全部亚型 / 单一亚型"的目标一致Target transcript ID + version recorded; target region matches the all-isoform vs single-isoform goal
- 靶区避开常见 SNP、重复区与同源基因区段Target region avoids common SNPs, repeat regions, and homologous-gene segments
- 候选共 3 条、位置分散（≥2 个外显子），每条 GC 在 40%–60%3 candidates, positions spread out (≥ 2 exons), each with GC 40%–60%
- 每条已写出 spacer 与 PAM，确认 PAM 不会被写进订购序列Spacer and PAM written out for each; confirmed the PAM will not be included in the order sequence
- 敲入项目：切口距编辑位点 ≤10 bp，donor 已加 PAM / seed 同义突变Knock-in projects: cut within ≤ 10 bp of the edit site; donor carries PAM / seed synonymous mutations
- 全基因组脱靶已查（≤4 错配），无全长 0–1 错配的第二位点；高风险位点有实验复查计划Genome-wide off-targets checked (≤ 4 mismatches); no second site with 0–1 mismatches genome-wide; high-risk sites have an experimental re-check plan
- 用 U6 表达的已排除 TTTT；化学合成的已选好修饰规格U6-expressed: TTTT excluded; chemically synthesized: modification spec chosen
- Cas9 形式已定（野生型 / 高保真 / nickase），与递送方式配套Cas9 form fixed (wild-type / high-fidelity / nickase), matched to the delivery method
- 阳性对照 sgRNA、阴性对照已一并订购Positive-control and negative-control sgRNAs ordered together
- 验证方案已定：引物不覆盖切口、TIDE + 单克隆测序 + Western 同步验证Validation planned: primers clear of the cut; TIDE + single-clone sequencing + Western together
| 参数Parameter | 常用值Typical value |
|---|
| spacer 长度Spacer length | 20 nt（SpCas9） |
| PAM | NGG（SpCas9）；变体见第 5 章variants in Chapter 5 |
| GC 含量窗口GC window | 40%–60% |
| 切口位置Cut position | PAM 上游约 3 bp~3 bp upstream of the PAM |
| 切口距敲入位点Cut-to-knock-in-site distance | ≤10 bp 为佳preferred |
| 验证 PCR 产物Validation PCR product | 约 400–800 bp，切口居中~400–800 bp, cut centered |
| 收样时间Harvest time | 递送后 48–72 h48–72 h after delivery |
| T7E1 检测限T7E1 detection limit | 约 1%–5%，仅作初筛~1%–5%, prescreen only |
| 敲除完成标准Knockout completion standard | 单克隆纯合移码 + 蛋白消失 + 表型验证Homozygous frameshift in single clones + protein loss + phenotypic validation |
| English | 中文Chinese | 一句话说明In one line |
|---|
| sgRNA | 单向导 RNASingle guide RNA | 人工合并 crRNA 与 tracrRNA 的向导分子Guide molecule fusing crRNA and tracrRNA |
| spacer | 间隔序列Spacer | sgRNA 中与靶 DNA 配对的 20 ntThe 20 nt of the sgRNA that pairs with target DNA |
| PAM | 原间隔序列邻近基序Protospacer adjacent motif | Cas9 识别的短基序，如 NGG；在基因组上Short motif Cas9 recognizes, e.g. NGG; on the genome |
| seed region | 种子区Seed region | spacer 中紧邻 PAM 的约 10–12 ntThe ~10–12 nt of the spacer next to the PAM |
| DSB | 双链断裂Double-strand break | Cas9 在切口处切断 DNA 双链Cas9 cuts both DNA strands at the cut site |
| NHEJ | 非同源末端连接Non-homologous end joining | 易错修复，产生随机 indel，敲除靠它Error-prone repair producing random indels; knockout relies on it |
| HDR | 同源定向修复Homology-directed repair | 按 donor 模板精确修复，敲入靠它Precise repair from the donor template; knock-in relies on it |
| indel | 插入 / 缺失Insertion / deletion | 切口修复后的小片段长度变化Small length changes after cut repair |
| off-target | 脱靶Off-target | 在非目标位点发生的切割Cutting at a non-target site |
| RNP | 核糖核蛋白复合物Ribonucleoprotein complex | 预组装的 Cas9 + sgRNA 递送形式Pre-assembled Cas9 + sgRNA delivery form |
| donor template | 供体模板Donor template | 敲入时提供的带同源臂的 DNA 模板DNA template with homology arms supplied for knock-in |
| high-fidelity Cas9 | 高保真 Cas9High-fidelity Cas9 | 脱靶更低的工程化 Cas9 变体Engineered Cas9 variant with lower off-targets |
| nickase | 切口酶Nickase | 只切单链的 Cas9 突变体，需成对使用Cas9 mutant cutting one strand only; used in pairs |
| TIDE / ICE | 测序分解工具Sequencing decomposition tools | 从混合 Sanger 峰图算 indel 比例Calculate indel percentages from mixed Sanger traces |
主要来源（可自行查阅原文）：Doench et al., Nature Biotechnology 2016（Rule Set 2 活性规则）；Hsu et al., Nature Biotechnology 2013（脱靶与 CFD 基础）；Kleinstiver et al., Nature 2016（SpCas9-HF1）；Chen et al., Nature 2017（HypaCas9）；Vakulskas et al., Nature Medicine 2018（HiFi Cas9）；Concordet & Haeussler, Nucleic Acids Research 2018（CRISPOR）。数据库与工具：NCBI RefSeq、Ensembl、dbSNP、Cas-OFFinder。本书为方法学教育资料，不构成对任何具体产品或序列的推荐。文中演示序列均为教学构造示例，不是验证过的产品，下单与实验前请按最新基因组版本与所在实验室规范复核。本书由 Fern Biotech 实验室指南系列出品。Main sources (look up the originals yourself): Doench et al., Nature Biotechnology 2016 (Rule Set 2 activity rules); Hsu et al., Nature Biotechnology 2013 (off-targets and CFD basis); Kleinstiver et al., Nature 2016 (SpCas9-HF1); Chen et al., Nature 2017 (HypaCas9); Vakulskas et al., Nature Medicine 2018 (HiFi Cas9); Concordet & Haeussler, Nucleic Acids Research 2018 (CRISPOR). Databases and tools: NCBI RefSeq, Ensembl, dbSNP, Cas-OFFinder. This guide is educational material on method and recommends no specific product or sequence. All demonstration sequences are teaching constructs, not validated products — re-verify against the current genome build and your lab's rules before ordering or experimenting. From the Fern Biotech Lab Guide Series.