A novel integrated strategy combining feature-based molecular networking, QSIIR modeling, and in silico toxicity...

As concealment methods for illegal cosmetic adulteration advance, identifying non-library or unexpected additives remains challenging. This study presents a novel integrated strategy for rapid, non-targeted detection of reference-standard-free illegal cosmetic additives, especially emerging structural analogs of known prohibited compounds. The quinolone case study validated the strategy's broad applicability to cosmetics. Mass spectrometry-based feature-based molecular networking (FBMN) leverages spectral similarity to cluster structural analogs, using known compounds to identify unknowns, thereby significantly expediting unknown compound discovery. Our results suggested that FBMN clustered 51 quinolones (including 14 novel synthetic analogs) into 13 structural groups using only 17 "seed" standards, achieving high sensitivity with a limit of detection (LOD) of 1 ppm. The quantitative structure-ionization intensity relationship (QSIIR) model was established using multiple linear regression (MLR). It employed 7 structural descriptors to predict concentrations of quinolones with high accuracy (training R² = 0.9876; prediction R² = 0.8221), enabling absolute quantification prediction of illegal quinolone additives in complex cosmetics without compound-specific calibrants. In addition, the OECD QSAR Toolbox software profiling identified consistent skin irritation hazards and structural alerts for genotoxicity/carcinogenicity across quinolones, supporting regulatory decision-making. Collectively, this FBMN-QSIIR-QSAR strategy can not only address the challenges from unavailable reference standards, complex matrices and trace-level additions but also provide regulatory authorities with an integrated screening-to-risk-assessment solution, contributing to the standardized development of the cosmetic safety control.