Mangat Chand S, Bharat Amrita, Gehrke Sebastian S, Brown Eric D
M. G. DeGroote Institute for Infectious Disease Research and Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, ON, Canada.
M. G. DeGroote Institute for Infectious Disease Research and Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, ON, Canada McMaster High Throughput Screening Laboratory, Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, ON, Canada
J Biomol Screen. 2014 Oct;19(9):1314-20. doi: 10.1177/1087057114534298. Epub 2014 May 14.
High-throughput screening (HTS) of chemical and microbial strain collections is an indispensable tool for modern chemical and systems biology; however, HTS data sets have inherent systematic and random error, which may lead to false-positive or false-negative results. Several methods of normalization of data exist; nevertheless, due to the limitations of each, no single method has been universally adopted. Here, we present a method of data visualization and normalization that is effective, intuitive, and easy to implement in a spreadsheet program. For each plate, the data are ordered by ascending values and a plot thereof yields a curve that is a signature of the plate data. Curve shape characteristics provide intuitive visualization of the frequency and strength of inhibitors, activators, and noise on the plate, allowing potentially problematic plates to be flagged. To reduce plate-to-plate variation, the data can be normalized by the mean of the middle 50% of ordered values, also called the interquartile mean (IQM) or the 50% trimmed mean of the plate. Positional effects due to bias in columns, rows, or wells can be corrected using the interquartile mean of each well position across all plates (IQMW) as a second level of normalization. We illustrate the utility of this method using data sets from biochemical and phenotypic screens.
对化学和微生物菌株库进行高通量筛选（HTS）是现代化学和系统生物学不可或缺的工具；然而，HTS数据集存在固有的系统误差和随机误差，这可能导致假阳性或假阴性结果。存在几种数据归一化方法；然而，由于每种方法都有局限性，没有一种方法被普遍采用。在这里，我们提出了一种数据可视化和归一化方法，该方法有效、直观且易于在电子表格程序中实现。对于每个平板，数据按升序排列，其绘图会生成一条曲线，该曲线是平板数据的特征。曲线形状特征直观地显示了平板上抑制剂、激活剂和噪声的频率和强度，从而能够标记出可能存在问题的平板。为了减少平板间的差异，可以通过排序后中间50%值的平均值对数据进行归一化，也称为四分位间距均值（IQM）或平板的50%截尾均值。由于列、行或孔中的偏差导致的位置效应可以通过所有平板中每个孔位置的四分位间距均值（IQMW）作为第二层归一化来校正。我们使用来自生化和表型筛选的数据集说明了该方法的实用性。
