抄袭Nature封面？加州大学教授实名举报中科院明星学者学术抄袭 后者正式回应

本文来源：公众号科研大匠、张老湿科研作图

7月3日下午，针对加州大学付向东教授实名举报中科院上海神经所80后明星教授杨辉学术抄袭、造假一事，中科院脑智卓越中心官网同时发表了中科院脑智卓越中心和当事人杨辉研究员的声明。

声明之后，杨辉对 BioArtReports 表示，付向东教授在神经所报告中提到的 Ptbp1 靶点，已经于 2013 年发表。关于我造假的言论，纯属污蔑。“他为了抹黑我，连基本的科学事实都不顾，这不是一个正派的科学家做的事情。”详见：

独家丨杨辉对付向东教授的回复

中心官网在上线两篇声明的同时，还同期上线了一篇Nature Methods（IF=30.822）5月19日对杨辉的一篇报道：Hui Yang_A better base editor with fewer off-target changes, from a die-hard Manchester United fan.（详见文末）

关注：加州大学付向东教授实名举报中科院上海神经所80后明星教授杨辉学术抄袭、造假？

声明中，中科院脑智卓越中心表示：对杨辉研究员相关舆情相当重视，已迅速成立由中心学术委员会成员和中科院外专家组成的调查组，对此事进行调查。将向社会反馈调查结果。

杨辉研究员声明：将如实提供原始材料，澄清事实，忠实履行一名科研工作者对科研诚信应尽的义务。

此前，网传杨辉曾对质疑作出正面回应，以下为网传杨辉之前的回应（尚未经本人证实，仅供参考对比）：

这是之前几天回复质疑的信，抱歉，附件和原邮件不能给你，response还在审稿中。对我每篇学术文章的质疑，欢迎发信给杂志社或者我来讨论，我只会回复学术的部分，谢谢！

关于我在MIT的文章，Genome Biology 文章(即质疑文章)刚在线的时候，我已经和editor取得联系，她欢迎我们写response回复，因为涉及到一些实验，最近刚刚和文章共同第一作者 Haoyi Wang与通讯作者Rudolf Jaenisch准备好，文章正在投稿中。以下是回复，详见附件：

Gurumurthyet al. [1]recently reported that a method developed by Yang et al. togenerate floxed allele (designated as “two donor method” by Gurumurthyet al.) [2] had poor reproducibility. They claimed that three centerscould not reproduce our results on generating conditional alleles of the Mecp2 locus and that the “two-donor method” had very low successful rate onother loci.

Here, weprovide our responses to these claims:

1. Our results on Mecp2 locus published byYang et al have been

reproducedby independent experiments in the Jaenisch (8-10% correct alleles ), Yang(8% correct alleles) and Hatada’s groups (2-6% correct alleles) [3] ,respectively.

2. The conditions used by Gurumurthy et al.[1] do not correspond to the conditionsused in our paper. The concentrations of CRISPR reagents used in the Gurumurthy et al.’s study [1]on the Mecp2 locus (10 ng/μl for Cas9 mRNA, 10ng/μl for sgRNA and 10 ng/μl for oligos) were much lower (10 fold lowerRNA and 20 fold lower oligo donor concentration) than those used in the Yang et al.’s experiments (Cas9 100 ng/μl, sgRNA 50 ng/μl and 100 ng/ μl foreach oligo) [2] and Yang et al.’s previous [4] and following publications[5-8]. It is well known that the concentrations of CRISPR reagentsare well correlated with the genome editing efficiency.

3. We utilized piezo-driven zygote injectionmethod in our original paper , whichallows for injecting CRISPR components at much higher concentration. The differencebetween this method and pronuclear injection method used by Gurumurthyet al. might also contribute to the difference of successful rates.

4. Multiple peer-reviewed publications [3,9-12] have successfully used ourmethod to create conditional knockout (CKO) mice (9 out of 11 loci succeeded,2.5% to 18% efficiency). We note that the efficiency of generating CKO mice by CRISPR/Cas9 is highly dependent on the professional skillsand well-built platforms. Thus, it is inappropriate to calculate the editingefficiency based on data from different labs and “core facilities” with varying capability and using different methodologies.

5. With any genome editing method or strategy being used, the

efficienciesat different genomic loci are often highly variable. In the

2013proof of concept paper, we showed the feasibility of generating floxed alleleat Mecp2 locus using CRISPR. As a X-linked gene, Mecp2 has a higher chance of having two independent loxP insertion events residing on the same Xchromosome, since half of the embryos are males. To assume the efficiency we demonstrated at Mecp2 locus will be directly translated to the successful rate atother genomic loci seems premature.

We agree with the Gurumurthy et al’s comment that the “one-donor method” offershigher success rate for generating floxed alleles in general, while the efficiency of “one-donor method” is also variable depending on the genomic loci and donor plasmid design. Before the publication of Gurumurthy et al., we also noted this, and developed a “one-donor method”, termed “Tild-CRISPR”method [8], and demonstrated the feasibility and high efficiencyin generating CKO mice.

With thefast improvement of genome editing technologies, we and many others constantly optimize our protocols. We welcome all discussions about the choiceof optimal strategy for particular applications, however, we think thereproducibility of any published work can only be validated by using the exactsame experimental methods and technical parameters.

PS: 和Genome Biology的editor 私下交流，这篇文章原本在Nature Methods审稿，但有一个reviewer就是用我们这种方法日常做CKO小鼠，所以文章拒了。转投GenomeBiology的时候，也有一个reviewer用我们这种方法日常做CKO小鼠，所以他们的文章关注点改为比较我们的CKO方法和最新发展的"one-donor method"，才发表出来。我一直不明白，许多人已经重复出我们的实验结果，也有许多人用该方法做CKO小鼠，这篇文章也能说造假吗？数据不可信吗？当然更多人做不出来，我认为很大原因是因为本身的实验体系不好，这篇文章的所有实验室（绝大大数是core facility）都没有先严格重复我们的实验结果，再在此基础上设计新的实验。我认为这不是受过好的科研训练的人应有的做事方式，我也没有收到他们的任何来信询问我该实验的实验细节和tricks。

关于我“其他文章也有很多问题”，我将我之前发表的文章都放到附件中，欢迎质疑者发信给我和大家，一一指出，我会做一一答复。

我也将我自博士期间以来发表的一些重要文章做一个简明阐述：

博士期间：Publication list 30: 孤雄单倍体的建立，这个已有一些实验室同样建立，比如周琪老师组，同时李劲松老师组也有许多后续的工作。但我相信更多实验室建立不了这个体系，技术要求太高，所以很难像CRISPR那样广泛使用。不过李劲松老师建立的孤雄单倍体平台会促进该技术的进一步应用。

博后期间：Publication list 27, 28: 首次报道利用CRISPR技术可以制作各种基因修饰小鼠，两篇文章都有数千次引用，也有无数个实验室重复和应用。当然技术还在不断优化和进步，我建立实验室后也发表过相应的文章，见Publication list14, 20.

独立PI期间：Publicationlist 18: 首次报道利用CRISPR可以敲除整体染色体。同期其他实验室也有报道，得到验证。

Publication list 7, 11：首次报道单碱基编辑技术的脱靶安全性问题，这两个工作都是back to back发表，相互很好的印证了彼此的结论。同时我们通过生物学改造获得高真单碱基编辑工具的文章也已经在Nature Methods接收，同期Nature Biotechnology的David Liu的文章的结果也跟我们有很好的印证。值得一提的是，我们在Science文章中建立的高敏感检测脱靶的GOTI技术，技术要求很高，只有少数实验室具备这个实验条件，但这些并不影响我们实验数据和结论的可靠性。

Publication list 1, 15：利用各种基因编辑技术做胶质细胞向神经元转分化研究，并且在最近的Cell文章中利用该方法治疗帕金森及RGC loss的小鼠模型。这篇文章两种疾病的治疗效果皆是该领域最好的，势必引起很大关注和争议，但我们相信在不久之后，许多实验室都能在自己各自的系统中得到验证。

最后说说我对最近一些事情的整体看法：

我个人对于外界的评论和质疑其实并不关注或看重，只要不是杂志编辑部或者同行邮件的询问质疑，都尽量置之不理。我非常珍惜现在中国良好的科研环境，所以格外珍惜每一点时间。只需一心在科研上面，不停有好的工作展示给同行，就足够了。也很钦佩像蒲老师、王晓东老师、施一公老师那样，呵护着我们这群青年PI，不用花费过多时间申请经费，有很好的core facility，在文章思路或写作上给予我们一定的指导。所有这些让我们有更多时间聚焦于科学本身。

此外，“最近一堆年轻人认为杨辉在MIT的文章涉嫌造假”。我相信绝大多数不是这个领域的专家，我认为正确的方式是写信给Cell的editor，质疑这篇文章的真实性、可靠性，而不是靠公众的舆论来挑取自己想要接收的信息。他们似乎不在乎质疑我们工作的文章的数据可靠性，也更接受10个月（其实不到半年时间）发两篇Cell只有造假才有的可能。然后近一年又把CNS都发了一遍，造假就实锤了。我觉得这不是一个好的风气，似乎许多人接受不了同代人比自己出色太多。

张锋一直是我追敢的目标，从PhD开始就一直很出色，很快的fellow经历，刚独立就发表了最重要的工作，后续更是一系列的重要文章，也成立了自己的公司，将CRISPR技术第一次应用于人治疗失明。

而我，PhD和Postdoc相对都很顺，但是回国独立头三年没有发表一篇文章，直到第5年才陆续有重要工作出现。也开始向临床迈进，希望两年内能够国内上临床，最终造福国内的患者。虽然时间有点落后张锋了，但我仍然坚信有赶上和超过他的一天。

所以希望““最近一堆年轻人”能找到自己追赶的目标，再为之努力，而不是传一些不在自己学术水平范围内能够judgement的事情。学术圈毕竟不是娱乐圈。

杨辉

以下为Nature Methods5月17日对杨辉的报道英文原文：

Hui Yang

A better base editor with fewer off-target changes, from a die-hard Manchester United fan.

When he was in high school in southern China, a teacher told him, “The 21st century is the century of life sciences,” says Hui Yang, a researcher at the Institute of Neuroscience at Shanghai Institutes for Biological Sciences, which is part of the Chinese Academy of Sciences. It’s why he chose to do his PhD research in biology. Yang studied at Shanghai Jiao Tong University and then completed his PhD research at the Shanghai Institute of Biochemistry and Cell Biology, focusing on developmental biology. He learned about gene editing while working on a project to generate androgenetic haploid stem cells, but he found the traditional strategies they used to be inefficient.

During a postdoctoral fellowship with Rudolf Jaenisch at the Whitehead Institute, Yang learned about CRISPR-based editing, and his first research project involved stem cells and reprogramming. It was nothing he had previously encountered in his textbooks, he says. “But I quickly fell in love with this.” With another postdoc, Haoyi Wang, he worked on generating gene-modified mice and found CRISPR powerful and easier to use than other approaches. In 2014 Yang was recruited back to China as part of the Youth Thousand Talents Program and started a lab. He shifted from creating genetically modified animal models to somatic editing focused on gene editing to one day treat human diseases.

Yang “is a talented young scientist fascinated by new technology,” says Mu-ming Poo, who directs the Institute of Neuroscience. “He has the audacity to pursue wild ideas, many of which began with a few sparks in his mind.” Yang fearlessly dives into the highly competitive field of gene editing, says Poo, in line with a Chinese saying: “a newborn calf is not afraid of the tiger.” Poo met Yang when the latter was a PhD student and they worked on a project involving gene manipulation in monkeys. After Yang returned from his postdoctoral fellowship, he joined the institute, where he enjoys Poo’s mentorship. “My hobby is just learning new things and seeking new challenges,” says Yang.

In his latest work, Yang and his team present base editor variants with improved fidelity and efficiency. He believes they have potential for clinical use, such as in human stem cells and ex vivo applications. But, for these and other clinical applications, “we need more accuracy,” he says. CRISPR–Cas-based editing usually involves double-stranded breaks. In base editing, which generally uses the rat cytidine deaminase APOBEC1 fused to the Cas9 nickase, only single stranded breaks are made. The result at the targeted site is conversion of cytosine to thymine. Yang and his colleagues have tuned base editors: they engineered the deaminase to achieve better expression and nuclear localization, says Yang. Their three base-editing variants generate fewer off-target changes and unwanted insertions and deletions than other base editors.

And they have a narrower base-editing window. “We do not want to correct one mutation and induce another mutation,” he says. But when cytosines are near the targeted base, bystander mutations can lead to unwanted on-target effects.

Using structure as a guide, the team focused on engineering the DNA-binding domain of the deaminase. The deaminase has its own DNA-binding domain, but it’s not needed; the Cas9 DNA-binding domain can be used instead, says Yang. They looked into which amino acids matter most, chose those, and screened for variants that affect DNA binding but that do not negatively affect the other desired traits: on-target efficiency and low off-target rates. Because the DNA-binding domain clusters with the RNA-binding domain, he believes the variants can be used for DNA and RNA editing. “Some amino acids are key for both,” he says. Some variants had higher off-target rates for RNA editing and others for DNA editing, but the researchers managed to find some with good efficiency and low off-target levels for both RNA and DNA editing.

For in vivo editing, he prefers RNA editing. The appeal is the smaller size of these editors and their higher efficiency. To him, RNA editing appears safer than DNA editing, although some concerns remain about non-specific cutting. “But I believe this could also be resolved,” he says.

For drug screening in which a broad editing window is acceptable, CRISPR is likely the right choice, says Yang. Base editing’s clinical promise is connected to the fact that many diseases are caused by point mutations, he says. He and a former student have founded a company called Hui-Gene Therapeutics to explore gene editing and human disease. Much work remains to be done, especially related to safety, he says.

“My hobby is just learning new things and seeking new challenges.”

Yang often gets ideas for the lab through interaction on social media, usually WeChat, but he prefers face-to-face discussion, he says. In the Jaenisch lab, he liked how valued independent thinking and interaction were and has styled his lab in that vein. Lab members can knock on his door anytime, he says. He helps students and postdocs and encourages more experienced lab members to guide others in designing and doing experiments. In lab meetings, “we just share ideas.” One of his students is in a joint program with a Danish university, and he wants to attract others from outside China to the lab. Once a week Yang plays basketball with friends, but he adores soccer. Since high school he’s been a die-hard Manchester United fan.

参考信息：

1.http://www.ion.ac.cn/sytzgg/202007/t20200703_5615873.html

2.http://www.ion.ac.cn/sytzgg/202007/t20200703_5615874.html

3.https://www.nature.com/articles/s41592-020-0857-1

本文来源：科研大匠整理自中科院脑智卓越中心官网，张老湿科研作图

本文不代表我方立场，仅用于学术分享。

“

文末说句正事

由于微信平台算法改版，公号内容将不再以时间排序展示，如果大家想第一时间看到我们的推送，强烈建议星标我们。具体步骤为：

（1）点击页面最上方“刘老师医学统计”，进入公众号主页

（2）点击右上角的小点点，在弹出页面点击“设为星标”，就可以啦。

感谢支持，比心

。（具体操作步骤见下方图片）

文章推荐ROC曲线还在用SPSS或Graphpad，就OUT了手把手教你用graphpad作图：重复测量方差分析的图形绘制【预测模型】构建COX回归模型并绘制列线图【预测模型】Logistic回归模型验证：绘制ROC曲线，计算AUC及95%CI，以及不同模型的比较【预测模型】单因素logistic回归，这么多自变量，重复点菜单，太浪费时间

温馨提示