A graph density-based strategy for features fusion from different peak extract software to achieve more metabolites in metabolic profiling from high-resolution mass spectrometry
A graph density-based strategy for features fusion from different peak extract software to achieve more metabolites in metabolic profiling from high-resolution mass spectrometry
复制标题
基于图密度的策略，用于来自不同峰提取软件的特征融合，以通过高分辨率质谱在代谢分析中获得更多代谢物
DOI：
10.1016/j.aca.2020.09.029
复制
发表时间：
2020-12-01
影响因子：
6.2
通讯作者：
Xu, Guowang
中科院分区：
文献类型：
--
作者：
Ju, Ran;Liu, Xinyu;Xu, Guowang
在代谢组学研究中，从超高效液相色谱 - 高分辨率质谱数据中提取代谢物并非易事，尤其是对于那些低丰度的代谢物。不同的峰识别与匹配软件采用不同算法，导致提取结果各异。因此，整合不同软件的结果能够获取更丰富的代谢组信息，但冗余特征需予以去除。在本研究中，提出了一种基于图密度的特征融合与去冗余集成策略（FRRGD）。利用图来涵盖两款开源软件（XCMS、MZmine 2）以及仪器供应商提供的一款软件（SIEVE）所生成的离子特征，并通过搜索最大完全子图来去除冗余特征。采用含有41种代谢物的标准混合物以及一份自发排尿样本来开发该方法并验证其有效性。对于标准混合物，XCMS、MZmine 2和SIEVE分别提取出19种、19种和27种代谢物。经FRRGD融合后，得到37种代谢物。对于稀释后的自发排尿样本，XCMS、MZmine 2和SIEVE分别提取出1103种、1500种和387种代谢物，而FRRGD得到1619种代谢物，远多于单个软件的提取数量，显著提高了代谢组覆盖范围。所提出的FRRGD作为代谢组学研究的一种新型数据处理策略展现出广阔前景。（C）2020爱思唯尔出版社。保留所有权利。
In metabolomics study, it is not easy to extract the metabolites from data of ultra high-performance liquid chromatography-high-resolution mass spectrometry, especially for those with low abundance. Different software for peak recognition and matching use different algorithms, leading to different extract results. Therefore, integration of results from different software can obtain richer metabolome information, but the redundant features should be removed. In this study, an integrated strategy of fusing features and removing redundancy based on graph density (FRRGD) was proposed. A graph is used to cover the ion features generated by two open access software (XCMS, MZmine 2) and a software (SIEVE) from an instrument vendor, and redundant features were removed by searching the maximal complete sub-graphs. A standard mixture containing 41 metabolites and a spontaneous urine were utilized to develop the method and demonstrate its usefulness. For the standard mixture, 19, 19 and 27 metabolites were extracted by XCMS, MZmine 2 and SIEVE, respectively. After fusion by FRRGD, 37 metabolites were obtained. For the diluted spontaneous urine sample, 1103, 1500 and 387 metabolites were extracted by XCMS, MZmine 2 and SIEVE, respectively, FRRGD produced 1619 metabolites which were much more than individual software, significantly increasing metabolome coverage. The proposed FRRGD shows a great prospect as a new data processing strategy for metabolomics study. (C) 2020 Elsevier B.V. All rights reserved.