For decades, environmental studies have looked to targeted analysis methods for accurate quantification of chemical exposures. The applicability of targeted analysis methods is limited to chemicals for which there are readily obtainable analytical standards, thus inhibiting characterization of chemicals of emerging concern for which standards may not be readily available. To complement targeted methods, non-targeted analysis (NTA) methods are now commonly used to identify tens to hundreds of chemicals in any sample of interest without prior knowledge of chemical presence within the sample. While NTA chemical identification methods are rapidly maturing, quantitative NTA (qNTA) methods have been somewhat slower to progress. Several qNTA methods are now able to generate point concentration estimates for identified chemicals, but none of these methods fully characterize the uncertainty of their estimates (<i>e.g.</i>, with confidence intervals). A clear need therefore exists for methods to bound chemical concentration estimates if NTA is to be utilized within regulatory contexts to support provisional risk characterizations. Here, the mathematical basis of traditional calibration methods used in targeted analyses is detailed, as well as procedures to extend these methods to support quantitative estimates for all chemicals observed in an NTA experiment. A non-parametric univariate method is detailed for bounding concentration estimates based on bootstrapping a distribution of instrument response factors for observed chemicals. A more complex method is then described, which utilizes predicted ionization efficiencies (calculated from machine-learning models using physicochemical descriptors) and linear mixed-effects models. The prediction uncertainty associated with both qNTA methods is compared to the uncertainty observed when performing semi-automated NTA experiments with traditional calibration curve-based approaches. We show that, in lieu of analytical standards, qNTA confidence limit estimation is feasible using either method given a base set of empirical response factors.

Estimating uncertainty of predicted chemical concentrations via quantitative non-targeted analysis

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.