Nontargeted Screening of Food Matrices: Development of a Chemometric Software Strategy To Identify Unknowns in Liquid Chromatography-Mass Spectrometry Data

The ability to identify contaminants or adulterants in diverse, complex sample matrixes is necessary in food safety. Thus, nontargeted screening approaches must be implemented to detect and identify unexpected, unknown hazardous compounds that may be present. Molecular formulas can be generated for detected compounds from high-resolution mass spectrometry data, but analysis can be lengthy when thousands of compounds are detected in a single sample. Efficient data mining methods to analyze these complex data sets are necessary given the inherent chemical diversity and variability of food matrixes. The aim of this work is to determine necessary requirements to successfully apply data analysis strategies to distinguish suspect and control samples. Infant formula and orange juice samples were analyzed with one lot of each matrix containing varying concentrations of a four compound mixture to represent a suspect sample set. Small molecular differences were parsed from the data, where analytes as low as 10 ppb were revealed. This was accomplished, in part, by analyzing a quality control standard, matrix spiked with an analytical standard mixture, technical replicates, a representative number of sample lots, and blanks within the sample sequence; this enabled the development of a data analysis workflow and ensured that the employed method is sufficient for mining relevant molecular features from the data.