Evidence level: C
一个基因同时捏了两只手:一只捏出你的斗型纹,一只捏你的造血干细胞。前者是「看得见的账单」,后者可能是真正的账本——本文设计一个中介检验,把这条通道拆开算账。
正文
前几篇我们讨论过 EVI1 的「一鱼两吃」:2022 年《Cell》的指纹 GWAS 把斗型纹的遗传信号定到了 EVI1 等肢体发育基因上(PMID 34995520),而同一个 EVI1 又是造血干细胞自我更新的核心调节因子,异常高表达是急性髓系白血病(AML)的强驱动子(PMID 19385966)。于是有一个自然的问题:斗型纹频率与白血病风险的观察性关联(1969 年起零散报道、结论互相矛盾,PMID 5112119),如果里面有真信号,它到底走哪条路?
答案是设计一个「垂直多效性中介检验」——用两步孟德尔随机化(MR)把总效应拆成三段:
第一段:暴露=斗型纹相关 SNP,中介=造血中间表型。 造血中间表型的工具变量现成:56 万欧洲人 + 74.6 万跨族裔的血细胞计数 GWAS 已把 29 种血液表型的遗传架构画到 5106 个独立变异(PMID 32888494),跨族裔版本又补了 5552 个关联(PMID 32888493)。用斗型纹 SNP(PMID 34995520 的汇总统计)做暴露,血细胞计数/网织红/血小板分布做中介,白血病结局(AML/MDS/ALL 分层)做终点,跑两样本两步 MR——这套流程在 2024-2025 年已被多个研究标准化,例如用血细胞计数中介肠道菌群→血压(PMID 38762013)、用免疫细胞中介血细胞扰动→神经病理性疼痛(PMID 39976904)。
第二段:算中介比例。 总效应 = 直接效应 + 中介效应。按实践指南(PMID 26653405)报告中介比例及其置信区间,并区分三种判读:
– 中介比例高(比如 >50%)且两段估计都显著 → 支持「造血发育通道」:EVI1 胚系变异先改变造血干细胞动力学(血细胞计数是它的行为学代理),白血病是这条通道的下游事件;
– 中介比例低但总效应仍在 → 通道可能在别处(如指尖垫形态学共享的发育窗口),或存在未测中介;
– 两段估计方向打架 → 直接判为矛盾证据,不许硬圆。
第三段:敏感性分析。 中介 MR 最大的坑是水平多效性——这里恰好叫「垂直多效性」:同一个 SNP 同时影响暴露和结局,绕过中介。对策:E-Value 量化需要多大强度的未测混淆才能把结论翻掉(PMID 26653405;E-Value 原始方法见 PMID 28693043);MR-Egger/加权中位数查定向多效性;用 colocalization 确认暴露-中介两段的信号是同一个因果变异而非 LD 搭便车。
这套检验的科学价值在于把一个 1969 年以来的模糊观察(白血病孩子皮纹有异常?PMID 5112119 后来复现失败,认为无诊断价值)升级成一个可判伪的因果拆账:如果中介比例经多数据源复现仍显著大于零,斗型纹就不再只是「民间手相」,而是一个造血发育通道的可观测代理指标;如果拆完发现总效应本身站不住,这条观察性关联就可以正式收档。 最新 EVI1 功能研究(PMID 42241446)继续确认它是 AML 的可成药依赖,说明这条通道末端是有生物学实体的。
证据与局限
已验证:EVI1 同时参与指纹形成与造血干细胞维持,两条线各有独立 GWAS/功能证据(PMID 34995520、19385966、32888494);两步 MR 中介框架与中介比例报告规范成熟(PMID 38762013、39976904、26653405)。
部分验证:斗型纹-白血病的观察性关联本身证据薄弱且互斥——1969 年 Lancet 报告异常掌纹增多,1971 年 54 例对照研究复现失败(PMID 5112119);EVI1 胚系变异经造血中间表型到白血病的中介比例从未被实测,本文全部设计推演,无一实测数字。
待验证:① 用指纹 GWAS + 血细胞 GWAS + AML GWAS 汇总统计跑两步 MR,输出中介比例及 E-Value;② colocalization 区分垂直多效性与 LD 假中介;③ 亚组拆分(AML vs ALL、儿童 vs 成人)。本文证据等级 C:方法链成熟、效应本身零实测。
相关阅读
- EVI1 多效性通道双方法验证:coloc + 双样本 MR
- 皮纹-疾病共定位 PP.H4 的东亚 vs 欧洲 LD 面板敏感性
- 皮纹+PRS 联合模型增量 NRI 的信号谱定位
- 研究池子:生物频率密码
证据等级: C
参考文献: PMID 34995520, PMID 19385966, PMID 32888494, PMID 32888493, PMID 38762013, PMID 39976904, PMID 26653405, PMID 28693043, PMID 5112119, PMID 42241446L1 网络状态:该主题尚未进入 L1 对应网络(CORRESPONDENCE_NETWORK 与 EDGE_TABLE_science-yixue 中均无皮纹/EVI1/白血病/中介 MR 相关边),本文按「薄证据」规则最高只能标 C,中介比例为设计推演而非实测。
免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。
EVI1 germline variants mediate the vertical pleiotropic effect of trisomy 21 on leukemia risk via hematopoietic intermediate phenotypes
Evidence level: C
A single gene simultaneously grasps two hands: one shapes your whorl pattern, the other your hematopoietic stem cells. The former is the “visible bill,” while the latter may be the true ledger—this study designs a mediation test to break down and calculate this pathway.
Body
In previous articles we discussed the “one gene, two bites” of EVI1: the fingerprint GWAS in Cell (2022) anchored the genetic signal of whorl patterns to EVI1 and other limb-development genes (PMID 34995520), while the same EVI1 is also the core regulator of hematopoietic stem cell self-renewal, with aberrant overexpression driving acute myeloid leukemia (AML) strongly (PMID 19385966). This raises a natural question: if there is a true signal in the observed association between whorl pattern frequency and leukemia risk (scattered reports since 1969 with contradictory conclusions, PMID 5112119), which pathway does it take?
The answer is to design a “vertical pleiotropy mediation test” — using two-step Mendelian randomization (MR) to decompose the total effect into three segments:
Segment 1: Exposure = whorl-associated SNPs; Mediator = hematopoietic intermediate phenotype. Tools for the hematopoietic intermediate phenotype are readily available: GWAS of blood cell counts in 560,000 Europeans and 746,000 multi-ethnic individuals have mapped the genetic architecture of 29 blood phenotypes to 5,106 independent variants (PMID 32888494), with the multi-ethnic version adding another 5,552 associations (PMID 32888493). Using whorl SNPs (summary statistics from PMID 34995520) as exposure, blood cell count/reticulocyte/platelet distribution as mediator, and leukemia outcome (stratified AML/MDS/ALL) as endpoint, we run two-sample two-step MR — this workflow has been standardized across multiple studies in 2024–2025, such as using blood cell count as mediator for gut microbiota → blood pressure (PMID 38762013), and using immune cells as mediator for blood cell perturbation → neuropathic pain (PMID 39976904).
Segment 2: Calculate the proportion mediated. Total effect = direct effect + mediated effect. Following practical guidelines (PMID 26653405), report the proportion mediated and its confidence interval, distinguishing three interpretations:
– High proportion mediated (e.g., >50%) and both estimates significant → supports the “hematopoietic development pathway”: germline variation in EVI1 first alters hematopoietic stem cell dynamics (blood cell count serves as its behavioral proxy), and leukemia is a downstream event along this pathway;
– Low proportion mediated but total effect remains → the pathway may lie elsewhere (e.g., shared developmental windows reflected in fingertip pad morphology), or there are unmeasured mediators;
– Conflicting directions of the two estimates → directly judged as contradictory evidence, not forced into coherence.
Segment 3: Sensitivity analysis. The biggest pitfall of mediation MR is horizontal pleiotropy — here we call it “vertical pleiotropy”: the same SNP affects both exposure and outcome, bypassing the mediator. Countermeasure: E-Value quantifies how strong an unmeasured confounder must be to overturn the conclusion (PMID 26653405; original E-Value method see PMID 28693043); MR-Egger/weighted median checks directional pleiotropy; colocalization confirms that the exposure-mediator segments share the same causal variant rather than riding on linkage disequilibrium.
The scientific value of this test lies in upgrading a vague observation since 1969 (leukemia children have abnormal skin prints? PMID 5112119 later failed to replicate, deemed without diagnostic value) into a falsifiable causal accounting: if the proportion mediated remains significantly greater than zero across multiple data sources, whorl patterns are no longer just “folk palmistry,” but an observable proxy indicator of a hematopoietic development pathway; if decomposition reveals the total effect itself is untenable, this observational association can be formally archived. Latest functional studies of EVI1 (PMID 42241446) continue to confirm it as a druggable dependency in AML, indicating the end of this pathway has biological substance.
Evidence and Limitations
Verified: EVI1 is confirmed to participate in both fingerprint formation and hematopoietic stem cell maintenance, with each pathway supported by independent GWAS and functional evidence (PMID 34995520, PMID 19385966, PMID 32888494). The two-step Mendelian Randomization (MR) mediation framework and the reporting standards for mediation proportions are well-established (PMID 38762013, PMID 39976904, PMID 26653405).
Partially Verified: The observational association between whorl-type fingerprints and leukemia is weak and mutually inconsistent. A 1969 Lancet report noted an increase in abnormal palm prints, but a 1971 study of 54 controls failed to replicate this finding (PMID 5112119). Furthermore, the mediation proportion from germline EVI1 variation through hematopoietic intermediate phenotypes to leukemia has never been measured experimentally; all calculations in this paper are derived entirely from theoretical modeling, with no empirical data points.
Pending Verification: ① Aggregate statistics from fingerprint GWAS, blood cell GWAS, and AML GWAS will be used to run the two-step MR analysis, outputting mediation proportions and E-Values. ② Colocalization analysis will distinguish vertical pleiotropy from linkage disequilibrium (LD) false mediation. ③ Subgroup splitting will be performed (AML vs ALL, children vs adults). This paper’s evidence level is C: the methodological chain is mature, but the effect itself has zero experimental measurement.
Related Reading
- EVI1 Multi-effect Pathway Dual-method Validation: coloc + Two-sample MR
- Dermatoglyphic-Disease Colocalization PP.H4 East Asian vs European LD Panel Sensitivity
- Dermatoglyphic + PRS Joint Model Incremental NRI Signal Spectrum Localization
- Research Pool: Bio-frequency Code
Evidence Level: C
References: PMID 34995520, PMID 19388494, PMID 32888494, PMID 32888493, PMID 38762013, PMID 39976904, PMID 26653405, PMID 28693043, PMID 5112119, PMID 42241446L1 Network Status: This topic has not yet entered the L1 correspondence network (no dermatoglyphic/EVI1/leukemia/mediator MR-related edges present in CORRESPONDENCE_NETWORK and EDGE_TABLE_science-yixue). Per the “thin evidence” rule, this is rated C at most, with mediator ratios derived from design inference rather than empirical measurement.
Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.
Leave a Reply