Machine learning driven decoding of impurity fingerprint in imidacloprid material

In recent years, imidacloprid, a first generation neonicotinoid pesticide, has attracted considerable attention due to its high detection rate in the environment and food, as well as its detrimental impact on human beings. The identification of structure-related impurities constitutes an essential aspect of pesticide registration and product standard development, and is of paramount importance for exposure risk and safety evaluation. In this study, the analysis of impurities in high-purity imidacloprid material was conducted using a combination of liquid chromatography-high resolution mass spectrometry (Orbitrap) and machine learning platform (SIRIUS + CSI:FingerID). The structure of IMI, as well as the feature fragments, were served as the foundation for decoding the chemical structures of the impurities, and this procedure was guided by the molecular formulas and fragment tree by SIRIUS + CSI:FingerID. Ultimately, 28 impurities were annotated and assigned confidence levels, of which 10 had CAS registry numbers, 5 had PubChem CIDs, and 4 were further verified by three dimension indexes (retention time, ultraviolet spectra and mass spectra) using standards. These impurities have been categorized into three levels (4 in level 1, 12 in level 2a and 12 in level 2b) according to their credibility. This approach was capable of providing accurate molecular formulas and possible structures of impurities, thereby greatly saving the time taken to resolve mass spectra. The elucidated impurity fingerprint can assist in the assessment of the potential health and ecological risks associated with imidacloprid and related impurities.