Reply to “Need for Rigor in Design, Reporting, and Interpretation of Transcriptomic Biomarker Studies”
Reply to “Need for Rigor in Design, Reporting, and Interpretation of Transcriptomic Biomarker Studies”
复制标题
DOI：
10.1128/jcm.06845-11
复制
发表时间：
2012-11
影响因子：
9.4
通讯作者：
Yurong Fu;Z. Yi;Xiaoyan Wu;Jianhua Li;Fuliang Xu
中科院分区：
文献类型：
--
作者：
Yurong Fu;Z. Yi;Xiaoyan Wu;Jianhua Li;Fuliang Xu
在生物医学研究中汇集样本已成为常见做法 (8,9,13)。例如，基因表达综合数据库中存放的数据集有 15% 来自混合 RNA 样本 (4)。支持合并策略的一个原因是合并生物样本的做法不仅减少了生物变异，而且还最大限度地减少了进行计算机密集型比较分析的数据文件的大小和数量 (7)。因此，共享策略仍然很有价值，并且在许多情况下都被推荐，特别是在资源有限的情况下。此外，最近的几篇出版物已经解决了合并样本可能隐藏生物学差异并对实验数据的推断意义产生错误信心的风险。例如，Kendziorski 等人。 (4)表明，即使生物平均不成立，合并也可能是有用的，并且关于差异基因表达的推论不会受到合并的不利影响。在我们最近的研究 (2) 中，使用合并策略来分析循环的 microRNA (miRNA) 群体，结果表明 miRNA 可能调节人类对结核分枝杆菌的免疫反应。为了减少生物学变异，所有受试者（结核病 [TB] 患者和健康对照之间的年龄和性别没有差异）都经过非常仔细的选择，并且所有样本都经过非常仔细的准备，以确保各个样本对池的贡献相同 (2)。我们的预测试结果表明，合并样本的 miRNA 谱总体上反映了单个样本。因此，为了减少生物变异和材料成本，我们的研究中采用了混合策略进行微阵列分析 (2)。此外，为了验证微阵列结果，使用对照和活动性结核病病例的个体样本进行定量实时逆转录 PCR (RT-PCR) 分析 (1, 10)。迄今为止，血浆 miRNA 非常稳定的内参尚未见报道 (5)，许多研究人员正在尝试寻找血清中 miRNA 表达的适当稳定的内参。尽管 miR-375 已被用作血浆中的内部对照 (5)，但在我们的研究中，结核病血清中的 miR-375 水平有所增加，因此不再将其作为适当的内部对照 (2)。 U6 之前也曾被用作血清对照 (3,6,14)。在我们的研究中，我们检测了U6表达，发现结核病病例和对照之间的U6水平没有差异（U6阈值周期[CT]水平，对照血清中为21.32，结核病血清中为21.29；这些数据代表n=30的平均值）。因此，U6 被用作血清中的内部对照，以标准化实时 PCR 分析中的 miRNA 表达 (2)。肺结核是一种危及生命的感染，通常伴有广泛的组织破坏和全身炎症状态。这些剧烈的生理变化无疑改变了活动性结核病患者的血清 miRNA 表达谱。我们研究的目的是鉴定活动性肺结核患者血清中与匹配健康对照血清中差异表达的 miRNA (11, 12)。此外，我们的结果最初表明，许多miRNA在活动性肺结核感染期间存在差异表达，并且基于这项单独的研究很难就参与活动性肺结核发病机制的miRNA得出明确的结论。因此，针对 Walter 博士及其同事在前一封信中提出的观点，我们得出的结论是，用于诊断结核病的生物标志物研究的完美设计可能永远不存在，没有完美的策略，有效挑战结核病标志物特异性的其他研究将非常有价值。需要进一步的长期研究来确定差异表达的 microRNA 是否代表宿主对结核病感染的特异性反应。
Pooling samples in biomedical studies has become frequent practice (8, 9, 13). For example, 15% of the data sets deposited in the Gene Expression Omnibus Database derive from pooled RNA samples (4). One reason in support of a pooling strategy is that the practice of pooling biological samples not only reduces biological variation but also minimizes the size and number of data files subjected to computer-intensive comparative analyses (7). A pooling strategy therefore remains valuable and is recommended in many settings, especially those which are resource constrained. Moreover, several recent publications have addressed the risk that pooling samples might potentially hide biological variance and give false confidence concerning the inferred significance of experimental data. For instance, Kendziorski et al. (4) showed that even when biological averaging does not hold, pooling can be useful and inferences regarding differential gene expression are not adversely affected by pooling. In our recent study (2), a pooling strategy was used in analyzing the circulating microRNA (miRNA) population and the results suggested that miRNAs likely regulate the human immune response to Mycobacterium tuberculosis. In order to reduce biological variation, all subjects (there were no differences in age and sex between tuberculosis [TB] patients and healthy controls) were very carefully selected and all samples were very carefully prepared to ensure that individual samples contributed equally to the pool (2). The results of our pretest indicated that the miRNA profiles of pooled samples in general mirrored individual samples. So, to reduce biological variation and material costs, a pooled strategy was used for microarray analysis in our study (2). Moreover, to validate the microarray results, individual samples from controls and active TB cases were used for quantitative real-time reverse transcription-PCR (RT-PCR) analysis (1, 10). Up to now, very stable internal controls of plasma miRNAs have not been reported (5), and many researchers are trying to discover appropriately stable internal controls for miRNA expression in serum. Although miR-375 has been used as an internal control in plasma (5), levels of miR-375 were increased in TB serum in our study, eliminating it as an appropriate internal control (2). U6 has also been used previously as a serum control (3, 6, 14). In our study, we detected U6 expression and found that there was no difference in U6 levels between TB cases and controls (U6 threshold cycle [CT] levels, 21.32 in control serum and 21.29 in TB serum; these data represent the average values of n 30). For this reason, U6 was used as the internal control in serum to normalize miRNA expression in real-time PCR analysis (2). Pulmonary TB is a life-threatening infection generally accompanied by extensive tissue destruction and a systemic inflammatory state. These drastic physiological changes undoubtedly alter the serum miRNA expression profile of patients with active TB. The goal of our study was to identify differentially expressed miRNAs in the sera of patients with active pulmonary TB in comparison to those in the sera of matched healthy controls (11, 12). Moreover, our results initially showed that a number of miRNAs were differentially expressed during active pulmonary TB infection and it was difficult to draw definitive conclusions as to the miRNAs involved in the pathogenesis of active pulmonary TB based on this single study. Accordingly, in response to the points raised by Dr. Walter and colleagues in the foregoing letter, we conclude that the perfect design of a biomarker study for the diagnosis of TB may never exist, no strategy is perfect, and additional studies that effectively challenge the specificity of TB markers would be exceedingly valuable. Further longer-term investigations are needed to determine whether differentially expressed microRNAs represent host responses that are specific or not to TB infection.