Abstract
A systematical study was performed to investigate the processing workflow of LC-MS-based metabolomics data by optimizing parameter settings in XCMS software and comparing different preprocessing methods. Here we use a spike-in experiment combining with design of experiment (DoE) approaches for optimizing XCMS software parameters. A trusted index, which was based on accuracy evaluation of the spike-in data, was employed to assess the optimizing process. After optimizing the XCMS setting, the trusted index was improved from 3.67 to 30 and positive rate of spike-in standards also increased from 20% to 100%. Moreover, different data preprocessing methods, such as normalization, different scaling methods were also investigated on spike-in data since they were found to affect the outcome of the data analysis and ions features identification. Accordingly, UN-normalization and Pareto scaling were chosen as appropriate preprocessing methods to deal with LC-MS data through the evaluation of match index (mainly applied multivariate statistics methods). Finally, the optimized workflow was applied to experimental samples that acquired from metabolomics experiment and analyzed randomly with spike-in sample, which indicated a better applicability in formal metabolomics experiment. It is concluded that the proposed data processing workflow could be used as feasible approach for improving the quality of LC-MS-based metabolomics data and ensured the veracity of metabolites identification in data processing procedures to a certain extent.