《计算系统生物学导论》系统生物学导论-2012-3.pptVIP

要点陈述 　研究背景 　研究内容概要 　总结和展望 芯片数据标准化 The source of noise Irregularities in the array surface Variations in the laboratory process Different scanner settings Different amounts of mRNA Different DNA strands have different hybridization properties Random noise 1. 图像分析 2. 扫描 3. DNA杂交过程 (温度、时间、混合均匀程度等) 4. 探针的标记 5. RNA的抽提 6. 加样 7. 其他 Data normalization supposed to adjust the global properties of measurements of individual samples so that they can be ‘‘more suitable compared’’ with a statistical benefit from removing large technical variations increasing the power of finding biological signals 传统标准化方法的假设 Quantile normalisation forces the probe intensities into the same distribution across all of the arrays. MAS5.0 normalises the probe intensities using linear scaling (based on a reference array) and brings the intensities of all of the arrays to the same level. DChip normalises the probe intensities using a rank-invariant set, forcing all of the arrays to have almost the same distribution of probe intensities as a baseline array 前提假设的质疑 研究内容分三个部分 一、基因在癌组织中差异表达的广泛性 二、癌相关表达、甲基化和拷贝数数据标准 化问题 三、cDNA芯片数据的可靠性分析 第一部分 基因在癌组织中差异表达的广泛性 验证同种癌型不同数据中找出的差异基因的表达改变方向的一致性。 推断出基因在癌组织中发生了广泛的差异表达。 表达的基因虽然在正常和疾病样本中的表达改变程度不一定显著，但是表达改变方向比较稳定。 第二部分 癌相关表达、甲基化和拷贝数数据标准化问题 Question ? 基因表达谱数据标准化问题 基因广泛差异表达的特征 Lowess Global Median center Quantile RMA MAS5.0 DChip Datasets pair-matched normal and cancer samples avoid the effects of certain complex factors, such as familial, individual and environmental differences collected all of the Affymetrix datasets according to the following criteria: each dataset must include at least 10 samples for each state (cancer or normal), all of the samples must have been generated on the same microarray platform. a total of 16 pair-matched datasets for 10 cancer types. 数据收集 配对样本数据集 批次效应 样本量20 The median of the PM probe intensities in the cancer samples increased in 14 out of the 16 dat