Samanipour Saer, Baz-Lomba Jose A, Alygizakis Nikiforos A, Reid Malcolm J, Thomaidis Nikolaos S, Thomas Kevin V
Norwegian Institute for Water Research (NIVA), Gaustadalléen 21, 0349 Oslo, Norway.
Norwegian Institute for Water Research (NIVA), Gaustadalléen 21, 0349 Oslo, Norway.
J Chromatogr A. 2017 Jun 9;1501:68-78. doi: 10.1016/j.chroma.2017.04.040. Epub 2017 Apr 20.
LC-HR-QTOF-MS recently has become a commonly used approach for the analysis of complex samples. However, identification of small organic molecules in complex samples with the highest level of confidence is a challenging task. Here we report on the implementation of a two stage algorithm for LC-HR-QTOF-MS datasets. We compared the performances of the two stage algorithm, implemented via NIVA_MZ_Analyzer™, with two commonly used approaches (i.e. feature detection and XIC peak picking, implemented via UNIFI by Waters and TASQ by Bruker, respectively) for the suspect analysis of four influent wastewater samples. We first evaluated the cross platform compatibility of LC-HR-QTOF-MS datasets generated via instruments from two different manufacturers (i.e. Waters and Bruker). Our data showed that with an appropriate spectral weighting function the spectra recorded by the two tested instruments are comparable for our analytes. As a consequence, we were able to perform full spectral comparison between the data generated via the two studied instruments. Four extracts of wastewater influent were analyzed for 89 analytes, thus 356 detection cases. The analytes were divided into 158 detection cases of artificial suspect analytes (i.e. verified by target analysis) and 198 true suspects. The two stage algorithm resulted in a zero rate of false positive detection, based on the artificial suspect analytes while producing a rate of false negative detection of 0.12. For the conventional approaches, the rates of false positive detection varied between 0.06 for UNIFI and 0.15 for TASQ. The rates of false negative detection for these methods ranged between 0.07 for TASQ and 0.09 for UNIFI. The effect of background signal complexity on the two stage algorithm was evaluated through the generation of a synthetic signal. We further discuss the boundaries of applicability of the two stage algorithm. The importance of background knowledge and experience in evaluating the reliability of results during the suspect screening was evaluated.
液相色谱-高分辨-四极杆飞行时间质谱（LC-HR-QTOF-MS）最近已成为分析复杂样品的常用方法。然而，在复杂样品中以最高置信度鉴定小分子有机化合物是一项具有挑战性的任务。在此，我们报告了一种针对LC-HR-QTOF-MS数据集的两阶段算法的实施情况。我们将通过NIVA_MZ_Analyzer™实施的两阶段算法的性能，与两种常用方法（即分别通过沃特世公司的UNIFI和布鲁克公司的TASQ实施的特征检测和提取离子流峰挑选）对四个进水废水样品进行可疑物分析的性能进行了比较。我们首先评估了通过两台不同制造商（即沃特世和布鲁克）的仪器生成的LC-HR-QTOF-MS数据集的跨平台兼容性。我们的数据表明，使用适当的光谱加权函数，两台测试仪器记录的光谱对于我们的分析物具有可比性。因此，我们能够对通过两台研究仪器生成的数据进行全光谱比较。对四种废水进水提取物分析了89种分析物，因此有356个检测案例。分析物分为158个经人工合成可疑分析物的检测案例（即通过目标分析验证）和198个真正的可疑物。基于人工合成可疑分析物，两阶段算法产生的假阳性检测率为零，而假阴性检测率为0.12。对于传统方法，假阳性检测率在UNIFI的0.06和TASQ的0.15之间变化。这些方法的假阴性检测率在TASQ的0.07和UNIFI的0.09之间。通过生成合成信号评估了背景信号复杂性对两阶段算法的影响。我们进一步讨论了两阶段算法的适用范围。评估了背景知识和经验在可疑物筛查过程中评估结果可靠性的重要性。
