Comparison of Software Tools for Liquid Chromatography-High-Resolution Mass Spectrometry Data Processing in Nontarget Screening of Environmental Samples
Comparison of Software Tools for Liquid Chromatography-High-Resolution Mass Spectrometry Data Processing in Nontarget Screening of Environmental Samples
复制标题
DOI：
10.1021/acs.analchem.9b04095
复制
发表时间：
2020-01-21
影响因子：
7.4
通讯作者：
Schmidt, Torsten C.
中科院分区：
文献类型：
--
作者：
Hohrenk, Lotta L.;Itzel, Fabian;Schmidt, Torsten C.
近年来，由于仪器的改进，导致仪器的灵敏度和选择性更高，高分辨率质谱领域经历了快速的发展。各种定性筛选方法，概括为非目标筛选，已被引入，并成功地扩展了环境监测的有机微污染物。已经开发了几种自动化数据处理工作流程，以处理通过这些方法在短时间内记录的大量数据。大多数数据处理工作流包括类似的步骤，但不同处理步骤的底层算法和实现方式各不相同。在这项研究中，数据处理的一致性与不同的软件工具进行了调查。为此，使用软件包MZmine2、enviMass、Compound Discoverer和XCMS在线处理相同的原始数据文件，并比较所得特征列表。结果显示，不同处理工具之间的一致性较低，因为所有四个程序之间的功能重叠约为10%，并且对于每个软件，40%至55%的功能与任何其他程序不匹配。重复和空白过滤器的实施被确定为观察到的差异的来源之一。然而，需要更好地理解不同算法和设置对特征提取和后续过滤步骤的影响，并提供用户说明。在未来的研究中，它将是感兴趣的调查如何最终的数据解释是由不同的处理软件的影响。通过这项工作，我们希望鼓励更多地认识到数据处理是非靶筛选工作流程中的关键步骤。
The field of high-resolution mass spectrometry has undergone a rapid progress in the last years due to instrumental improvements leading to a higher sensitivity and selectivity of instruments. A variety of qualitative screening approaches, summarized as nontarget screening, have been introduced and have successfully extended the environmental monitoring of organic micropollutants. Several automated data processing workflows have been developed to handle the immense amount of data that are recorded in short time frames by these methods. Most data processing workflows include similar steps, but underlying algorithms and implementation of different processing steps vary. In this study the consistency of data processing with different software tools was investigated. For this purpose, the same raw data files were processed with the software packages MZmine2, enviMass, Compound Discoverer, and XCMS online and resulting feature lists were compared. Results show a low coherence between different processing tools, as overlap of features between all four programs was around 10%, and for each software between 40% and 55% of features did not match with any other program. The implementation of replicate and blank filter was identified as one of the sources of observed divergences. However, there is a need for a better understanding and user instructions on the influence of different algorithms and settings on feature extraction and following filtering steps. In future studies it would be of interest to investigate how final data interpretation is influenced by different processing software. With this work we want to encourage more awareness on data processing as a crucial step in the workflow of nontarget screening.