Measurable Feature Prediction for Estimating Chemical Space Coverage in LC-ESI-HRMS Non-targeted Analysis

The chemical space, encompassing both known and theoretically possible compounds, continues to expand thanks to anthropogenic activity and transformation processes. Understanding which regions of this vast space are experimentally measurable is critical for advancing non-targeted analysis (NTA) for exposomics and environmental monitoring. In practice, comprehensive NTA by LC–ESI–HRMS is constrained by method-specific conditions, such as retention and ionization, thus defining subspace regions of measurable chemical entities. To address this limitation, we present a bottom-up in silico framework for predicting the measurable features and the fraction of chemical space accessed under specific LC–ESI–HRMS conditions, before sample analysis. The approach integrates experimental data from internal standards detected in the chromatographic and mass domains, with molecular fingerprints and quantitative structure–property relationships such as retention index and ionization efficiency derived from large-scale chemical databases (CompTox). Structural similarity and measurability predictions of retention index and ionization efficiency are achieved through distance metrics and optimized k-nearest neighbor regression modeling. Applied to internal standards and chemical space datasets, this framework identifies chemical neighbors amenable to detection (i.e., what is measurable by the method) and provides quantitative estimates of the method-specific coverage. Additionally, the direct coverage comparison of different experimental setups highlights both shared and unique measurable chemical regions, demonstrating how adopting orthogonal methods can expand chemical coverage and diversity in NTA studies. The presented framework provides a chemical space scalable strategy for predicting chemicals amenable to detection, providing a better understanding of the measurability of chemicals in NTA.