Abstract
The growing diversity of anthropogenic chemicals in the environment far exceeds the scope of routine analytical monitoring. Non-target screening (NTS) using high-resolution mass spectrometry (HRMS) has thus emerged to discover unknown organic contaminants. Liquid or gas chromatography (LC/GC) coupled with ion mobility-mass spectrometry (IM-MS) further enhances NTS by providing multidimensional, structurally informative data. Machine learning (ML) offers a powerful solution by efficiently processing high-dimensional data and uncovering patterns. Both supervised and unsupervised learning approaches show strong potential to streamline labor-intensive processes. This review provides an overview of key ML algorithms and representative workflows in LC/GC-(IM-) MS-related NTS, followed by a critical synthesis of recent advances in ML-enabled applications across the entire NTS procedure, from sample analysis to data acquisition, and ultimately risk assessment. Continued advances in ML are expected to transform NTS into a more efficient and robust tool for risk assessment.