An integrated strategy for rapid discovery and prediction of nucleobases, nucleosides and amino acids as quality markers in different flowering stages of Flos Chrysanthemi using UPLC–MS/MS and FT-NIR coupled with multivariate statistical analysis

•UPLC–MS/MS and FT-NIR was integrated for rapid quality evaluation of Flos Chrysanthemi.•UPLC–MS/MS was to measure 40 nutrients of Flos Chrysanthemi at three flowering stages.•10 quality markers were picked to distinct Flos Chrysanthemi in three flowering stages.•Glutamate, asparagine and aspartate were key quality markers for three flowering stages.•PLSR models were constructed to quantify the three key quality markers based on FT-NIR. Nucleobases, nucleosides, nucleotides and amino acids are important nutritional compounds, and some of them are positively correlated with the umami taste in scented tea. Taking these nutrients of Flos Chrysanthemi from three different flowering stages as the research object, an integrated strategy of ultra high-performance liquid chromatography coupled with triple-quadrupole linear ion-trap tandem mass spectrometry (UPLC–MS/MS), Fourier transform near infrared spectroscopy (FT-NIR) and multivariate statistical analysis was proposed to set up a systematic quality evaluation system for Flos Chrysanthemi from discovery to the rapid prediction of quality markers. First, a rapid, sensitive and accurate UPLC–MS/MS method was established to simultaneously determine 40 nucleobases, nucleosides, nucleotides and amino acids in three different flowering stages of Flos Chrysanthemi. The results indicated that Flos Chrysanthemi from three different flowering stages are rich in 40 nutritional compounds detected above and could be developed as functional products possessed nutraceutical and bioactive properties. Next, ten potential quality markers for three flowering stages of Flos Chrysanthemi were screened using partial least squares discriminant analysis (PLS-DA) and the plot of variable importance for projection (VIP), and the related metabolic pathways were identified for the first time. Furthermore, three of the ten potential quality markers, namely, glutamate, asparagine and aspartate were selected as key quality markers. Finally, three best PLS regression models constructed with FT-NIR were developed to predict the contents of the three key quality markers using the optimal spectral preprocessing methods and spectral interval combinations screened by synergy interval-PLS (Si-PLS), which yielded excellent prediction performance: 0.9766 ≤ R2p (coefficient of determination for prediction) ≤ 0.9858 and 6.3059 ≤ RPD (ratio of prediction to deviation) ≤ 7.3613. To sum up, the integrated quality marker discovery and FT-NIR quantitation