An understanding of the contribution of each original variable to the synthetic variables leads to the identification of key variables that contribute to the relationships among samples. In this study, we carried out multivariate data analyses using SIMCA-P+ version 12.0. PCA models are depicted as score plots and consist of two synthetic variables: principal component 1 and PC2. These display intrinsic groups of samples based on spectral variations. The corresponding loading plots show the contribution of each spectral variable to score formation. Therefore, this analysis can explain the original feature of samples based on the ratio of the sum of percentages of PC1 and PC2. All variables obtained from LC-MS datasets were mean-centered and scaled to Pareto variance. The quality of OPLS-DA models was evaluated by the goodness-of-fit parameter R2 and the predictive ability parameter Q2. R2 and Q2 values higher than 0.5 indicated good quality of OPLS-DA models. Metabolite peaks were assigned by MS/MS analysis or by searching their accurate masses using online metabolite databases. PLS, PLS-orthogonal signal correction , and OPLS were chosen to create the prediction model. PLS, which can be described as the regression extension of PCA, was calculated using SIMCA-P+. PLS derives latent variables that maximize the covariation between measured metabolite data and the response variable regressed against. This differs from PCA, which utilizes the OTX015 Epigenetic Reader Domain inhibitor maximum variation in the metabolite data matrix. OSC is normally used to remove uncorrelated variables or those orthogonal to inhibitory activity from metabolite data using the nonlinear iterative partial least-squares algorithm. Aqueous crude extracts of tea leaves from the 43 cultivars were subjected to LC-MS to investigate differences in their compositions. In analyses of complex mixtures such as crude extracts, two or more compounds can be co-eluted. The obtained complex spectral data are usually processed to extract and align peaks. We extracted 541 peaks from a complex chromatogram and used multivariate statistical analysis to decrease the complexity of the spectra datasets. This chemometric approach has the potential for use in HhAntag691 classification and bioactivity assessment without any prepurification methods such as extraction of arbitrary constituents from crude extracts prior to LC-MS measurement.