Different software processing affects the peak picking and metabolic pathway recognition of metabolomics data

In untargeted liquid chromatography‒mass spectrometry (LC‒MS) metabolomics studies, data preprocessing and metabolic pathway recognition are crucial for screening important pathways that are disturbed by diseases or restored by drugs. Here, we collected high-resolution mass spectrometry data of serum samples from 221 coronary heart disease (CHD) patients under two different chromatographic columns (BEH amide and C18 column) and evaluated the three commonly used software programs (XCMS, Progenesis QI, MarkerView) from four aspects (including signal drift, peak number, metabolite annotation and metabolic pathway enrichment). The results showed that the data preprocessed by the three software programs have different degrees of signal drift, but the StatTarget could improve the data quality to meet the data analysis requirement after correction. In addition, XCMS surpassed other software in detection of real chromatographic peaks and Progenesis QI was the best performer in terms of the number of metabolite annotation. XCMS and Progenesis QI showed different performance in pathway enrichment. However, metabolic pathways based on the combination of XCMS and Progenesis QI had a high coincidence with Progenesis QI. In addition, we also reported that C18 and amide columns were highly complementary and have great potential for cooperation in the context of metabolic pathways. In this study, the effects of different chromatographic columns and software pretreatments on metabolomics data were evaluated based on clinical large cohort samples, which will provide a reference for the metabolomics of clinical samples and guide subsequent mechanistic research.