Exploring AI-enhanced NMR dereplication analysis for complex mixtures and its potential use in adulterant detection

Identifying chemical entities in complex mixtures is crucial for advancing research in natural products and metabolomics. This enables the discovery of unique structures and bioactive compounds. Dereplication analysis streamlines this process by rapidly identifying known compounds, reducing redundant efforts, and accelerating the exploration of bioactive entities. NMR is the most effective technique for the identification of small organic molecules due to its robust structural elucidation capabilities. However, challenges such as low sensitivity and complex sample matrices often lead to overlapping NMR signals and spectral ambiguities. Artificial intelligence (AI), especially deep learning, has revolutionized dereplication by significantly enhancing pattern recognition in applications ranging from regression and classification to clustering. This review explores the origins, development, and principles of dereplication techniques, as well as the features of mass spectrometry and NMR. It also delves into the advancements in AI-enhanced NMR dereplication, elucidating the principles for various models and the key issues they address during the dereplication process. Furthermore, it covers the methods for constructing reference databases, and the usability and accessibility of these models. These insights can assist researchers, including those new to AI algorithms, in rapidly adopting and implementing these advanced methods. Moreover, this review highlights the broader utility of dereplication methods in fields such as the detection of illegal additives in cosmetics, emphasizing their potential applications in a wider range of fields.