Abstract：
LC-MS based multi-attribute methods (MAM) have drawn substantial attention due to their capability of simultaneously monitoring a large number of quality attributes of a biopharmaceutical product. For successful implementation of MAM, it is usually considered a requirement that the method is capable of detecting any new or missing peaks in the sample when compared to a control. Comparing a sample to a control for rare differences is also commonly practiced in many fields for investigational purpose. Because MS signal variability differs greatly between signals of different intensities, this type of comparison is often challenging, especially when the comparison is made without enough replicates. In this report we describe a statistical method for detecting rare differences between two very similar samples without replicate analyses. The method assumes that an overwhelming majority of components have equivalent abundance between the two samples, and signals with similar intensities have similar relative variability. By analyzing several monoclonal antibody peptide mapping datasets, we demonstrated that the method is suitable for new-peak detection for MAM as well as for other applications when rare differences between two samples need to be detected. The method greatly reduced false positive rate without a significant increase of false negative rate.
摘要：
基于LC-MS的多属性方法(MAM)由于其同时监测生物制药产品的大量质量属性的能力而引起了大量关注。为了成功实施MAM，当与对照相比时,通常认为要求该方法能够检测样品中的任何新的或缺失的峰。为了研究目的，在许多领域中通常也实践将样品与对照进行罕见差异的比较。由于MS信号的变异性在不同强度的信号之间差异很大，这种类型的比较通常是具有挑战性的，特别是当没有足够的重复进行比较时。在本报告中，我们描述了一种统计方法，用于检测两个非常相似的样品之间的罕见差异，而无需重复分析。该方法假设绝大多数成分在两个样品之间具有相等的丰度，和具有相似强度的信号具有相似的相对变异性。通过分析几个单克隆抗体肽图谱数据集，我们证明了该方法适用于MAM的新峰值检测以及其他需要检测两个样品之间罕见差异的应用。该方法大大降低了假阳性率，而没有显着增加假阴性率。
