Automated correction of saturated peaks in untargeted GC-MS: a chemometric approach

Gas chromatography (GC) coupled with mass spectrometry (MS) is a widely utilized analytical technique in various research fields, such as food and environmental science to generate untargeted profiles of volatile organic compounds (VOCs). These profiles often contain numerous compounds with diverse functional groups, present in a wide range of concentrations. Despite the multiple advantages of this technique, highly concentrated compounds can lead to problems such as MS detector saturation. This results in a distorted signal and a loss of accurate quantitative information, as the peak area of the saturated ion (m/z) is no longer a reliable estimate of a compound's abundance in a sample. Although detector saturation is a common challenge in chromatography, no robust solution has been developed to effectively address it. Therefore, this study proposes an automated method to detect and correct saturated m/z signals in untargeted GC-MS data, enabling accurate quantification and identification of highly concentrated VOCs. Our approach leverages the fact that saturation in fact manifests as right-censored data in specific masses (m/z), allowing automatic detection and unbiased modelling using tensor decomposition, namely parallel factor analysis 2 (PARAFAC2) models, combined with expectation-maximization imputation to restore saturated m/z values.