Comprehensive phytochemical profiling and ensemble machine learning discrimination of frozen mango puree processed by different methods.

High-pressure processing (HPP) and high-temperature short-time (HTST) pasteurization are widely used to stabilize mango puree, but their effects on phytochemical composition remain unclear. Here, untargeted UHPLC-QTOF metabolomics using RPLC/HILIC, targeted UPLC-QQQ quantification, pseudotargeted MRM profiling, and ensemble learning were integrated to compare non-sterilized, HPP-, and HTST-treated mango purees. In total, 311 phytochemicals were annotated, including 51 compounds putatively reported in mango for the first time. Processing-related changes were mainly associated with phenolics and amino acid derivatives. Targeted quantification showed higher ferulic acid in HPP than HTST puree (0.61 vs 0.13 mg/100 g FW), while methyl gallate decreased markedly after HTST treatment. A pseudotargeted MRM panel selected nine discriminant markers, enabling robust classification of processing methods and preliminary authentication of commercial mango-based products. This study provides a practical metabolomic-chemometric framework for industrial quality control and processing traceability.