Weis S, Llenos I C, Dulay J R, Elashoff M, Martínez-Murillo F, Miller C L
Stanley Laboratory for Brain Research and Neuropathology, Department of Psychiatry, Uniformed Services University of the Health Sciences, and Stanley Medical Research Institute, Bethesda, MD, USA.
J Neurosci Methods. 2007 Sep 30;165(2):198-209. doi: 10.1016/j.jneumeth.2007.06.001. Epub 2007 Jun 7.
The quality of results from microarray studies depends on RNA quality, which can be significantly influenced by postmortem factors. The aim of this study was to determine which postmortem factors and/or RNA electropherogram characteristics best correspond to microarray output and can be used to prospectively screen RNA prior to microarray analysis. Total RNA was extracted (N=125) from gray and white matter of postmortem frontal and occipital lobe tissue, acquired from normal controls, and patients with schizophrenia, bipolar disorder or major depression. Electropherograms were generated by the Agilent BioAnalyzer 2100, allowing calculation of the 28S/18S ratio, the 18S/baseline peak ratio and the RNA Integrity Number (RIN). These values were compared to post-hybridization image analysis of Affymetrix microarrays. The postmortem variables correlated with some quality measures but could not be used as effective screening tools. Logistic regression demonstrated that all three electropherogram measures were predictive for microarray quality, and that the RIN threshold predictive of "good quality" (>35% present calls) was most consistent with that of prior studies. The optimal RIN must be determined by the investigator's specifications for false inclusion and false exclusion. In contrast to RIN, the quality threshold for the 28S/18S ratio has proven unacceptably variable, due to sensitivity to slight differences in protocol and/or tissue source. In conclusion, the measures we found useful as screening criteria do not replace the need to exclude samples after a microarray analysis is performed, as an acceptable percent call rate and other measures of microarray quality represent the desired endpoint.
微阵列研究结果的质量取决于RNA质量，而RNA质量会受到死后因素的显著影响。本研究的目的是确定哪些死后因素和/或RNA电泳图谱特征与微阵列输出最相符，可用于在微阵列分析前对RNA进行前瞻性筛选。从正常对照、精神分裂症患者、双相情感障碍患者或重度抑郁症患者死后的额叶和枕叶组织的灰质和白质中提取总RNA（N = 125）。通过安捷伦生物分析仪2100生成电泳图谱，从而计算28S/18S比值、18S/基线峰比值和RNA完整性数值（RIN）。将这些值与Affymetrix微阵列的杂交后图像分析结果进行比较。死后变量与一些质量指标相关，但不能用作有效的筛选工具。逻辑回归表明，所有这三种电泳图谱指标都可预测微阵列质量，且预测“高质量”（>35%有信号）的RIN阈值与先前研究最为一致。最佳RIN必须由研究者对假纳入和假排除的规定来确定。与RIN不同，28S/18S比值的质量阈值因对实验方案和/或组织来源的细微差异敏感，已被证明具有不可接受的变异性。总之，我们发现作为筛选标准有用的指标并不能取代在进行微阵列分析后排除样本的需求，因为可接受的信号检出率百分比和微阵列质量的其他指标才是期望的终点。
