Recent Advances in Implementation of Machine Learning for Environmental Nontarget Identification

The continuous emergence of new pollutants poses significant threats to both human health and ecological environments. Nontarget analysis based on mass spectrometry has become prevalent for detecting new pollutants due to its high throughput capabilities. However, structural elucidation remains a major challenge in nontarget analysis. Here, we review the implementation of machine learning techniques to accelerate nontarget structural elucidation, with particular focus on spectral library matching, structural database retrieval, and de novo structure generation. We investigated the design principles, technical characteristics, and comparative evaluation of these computational approaches. In addition, we show their applications in environmental nontarget analysis for new pollutant identification. Finally, we discuss the challenges of current approaches and the future development trends. This review aims to deepen the understanding of existing computational approaches, promote the application of machine learning techniques in nontarget identification, and facilitate the integration of artificial intelligence with environmental pollutant analysis.