MARS 2: A computational tool to resolve and extract features from large-scale GC-MS datasets

Abstract

MARS (MS-Assisted Resolution of Signal) is a computational tool for feature extraction using multivariate curve resolution approaches. However, it is time-consuming for large-scale GC-MS datasets and sensitive to the component number estimation. Here we introduce MARS 2 with substantial improvements to overcome these limitations. Reverse matching (RM) and principal component optimization-iterative transformation target factor analysis (PCO-ITTFA) are developed to reduce the time for locating peak regions and determine the correct component number for resolving optimal features. The greatest strength of MARS 2 is “resolve once, extract anywhere”. It means that mass spectra of each component were needed to resolve from only one profile. Then, chromatographic features can be extracted automatically from hundreds of other profiles with the assistance from the resolved mass spectra. To evaluate its performance, plasma datasets and fatty acid standard mixtures were processed by MARS 2, AMDIS, ADAP-GC 3.0, eRah and MS-DIAL. Amino acid standard mixtures were analyzed by MARS 2 for calibration. Results show that MARS 2 can achieve better performance in both qualitative and quantitative analysis. It is implemented in Python programming language and open-sourced at https://github.com/mapancsu/MARS2.