Using structural similarity to estimate concentrations of known unknowns in suspect screening analyses

摘要

Suspect screening analysis (SSA), based on high-resolution mass spectrometry (HRMS), has recently been adopted in environmental chemistry applications to enable broad-scale chemical surveillance and exposure monitoring. SSA is inherently a qualitative technique, and generally unable to provide quantitative information for a given compound in the absence of a conventional reference standard. Here, we introduce novel methodology that enables quantitation of known unknowns (i.e., HRMS features for which a probable structure can be put forth via identification algorithms and a custom SSA workflow) in samples without the use of conventional standards. Three blinded synthetic mixtures, based on US EPA ToxCast chemicals, were analyzed using liquid chromatography quadrupole time-of-flight mass spectrometry. Calibration curves were generated for tentatively identified compounds (via SSA), in the first two mixtures. SSA was then performed on the third mixture at multiple blinded dilutions. Molecular formulae were assigned to features in the third mixture and mapped to candidate structures using the US EPA's CompTox Chemistry Dashboard. Ranking/classification tools, along with novel semi-Data Dependent MS/MS methods, were used to propose and corroborate chemical structures for each formula. Chemical similarity coefficients were used to link candidate compounds from the third mixture to structurally-similar compounds with existing calibration data from the first two mixtures. Quantitation for each candidate compound, at each dilution, was performed using surrogate calibration curves, with a weighting scheme based on chemical similarity score (i.e., structurally-similar compounds were given more weight). The techniques described here allow for semi-quantitation of unknowns detected by SSA, and thus provide a means to generate exposure and dosage information to be considered in support of avant-garde chemical screening programs.