Comparative evaluating laser ionization and iKnife coupled with rapid evaporative ionization mass spectrometry and machine learning for geographical authentication of Larimichthys crocea

Larimichthys crocea (LYC) holds significant economic value as a marine fish species. However, inaccuracies in labeling its origin can adversely affect consumer interests. Herein, a laser assisted rapid evaporative ionization mass spectrometry (LA-REIMS) and machine learning (ML) was developed for geographical authentication. When compared to iKnife, the LA demonstrated to be superior owing to reduced thermal damage to sample tissue, enhanced automation, and ease of use. Analysis of LYC from six distinct geographical origins across China revealed a total of 798 ions, which were then subjected to six classifiers to establish ML models. Following hyperparameter optimization and feature engineering, the Chi2(15%)-KNN model exhibited the highest training and testing accuracy, achieving 98.4 ± 0.9% and 98.5 ± 1.4%, respectively. This LA-REIMS/ML methodology offers a rapid, accurate, and intelligent solution for tracing the origin of LYC, thereby providing valuable technical support for the establishment of traceability systems in the aquatic product industry. •A LA-REIMS method was developed for geographical authentication of LYC.•Lipid phenotypic differences of Larimichthys crocea from six origins was analyzed.•LA-REIMS demonstrated less tissue thermal damage compared to iKnife-REIMS.•ML-guided LA-REIMS pattern recognition enabled accurate and intelligent analysis.