A novel integrated strategy combining feature-based molecular networking, QSIIR modeling, and in silico toxicity prediction accelerates the screening of illegal additives in cosmetics: Quinolones as a case study.

Summary

This analytical workflow integrates FBMN, a QSIIR-MLR model, and in silico toxicity prediction to non-targetedly detect and prioritize illegal quinolone adulterants in cosmetics. It clustered 51 quinolones (14 novel) into 13 groups with 1 ppm LOD using only 17 seed standards and accurately predicted concentrations from structural descriptors.

Key Findings

• FBMN clustered 51 quinolones, including 14 novel analogs, into 13 structural groups using 17 seed standards.
• Achieved a limit of detection of 1 ppm for quinolones in cosmetic contexts.
• Developed a QSIIR model (MLR with 7 structural descriptors) to accurately predict quinolone concentrations.
• Integrated in silico toxicity prediction to prioritize potentially hazardous adulterants.

Clinical Implications

Enhances public health by enabling earlier detection of harmful adulterants, informing recalls, and guiding toxicology follow-up; supports cosmetic dermatologists in identifying exposure sources in adverse reactions.

Limitations

• Validation reported for quinolones; generalizability to other prohibited classes requires demonstration.
• External, real-world surveillance datasets and inter-laboratory reproducibility were not detailed.

Future Directions

Expand to other adulterant classes, prospectively validate across laboratories and product matrices, and link analytical flags to clinical toxicovigilance outcomes.