Abstract
In the process of anticancer drug development, research on the mechanism of action remains a major obstacle. In the present study, a cell metabolic profiling based discriminatory model was designed to give general direction on anticancer candidate mechanisms. Firstly, ultra-performance liquid chromatography in tandem with high-definition mass spectrometry was applied to obtain a comprehensive metabolic view of 12 human tumor cells. Secondly, multivariate data analysis was used to assess the metabolites' variations, and 42 metabolites were identified as the main contributors to the discrimination of different groups. Then a metabolite-based prediction model was constructed for the first time and verified by cross validation (R (2) = 0.909 and Q (2) = 0.869) and a permutation test (R (2) = 0.0871 and Q (2) = -0.4360). To validate if the model can be applied for mechanism prediction, 4 independent sample sets were used to train the model and the data dots of different drugs were located in different regions. Finally, the model was applied to predict the anticancer mechanism of two natural compounds and the results were consistent with several other studies. Overall, this is the first experimental evidence which reveals that a metabolic profiling based prediction model has good performance in anticancer mechanism research, and thus it may be a new method for rapid mechanism screening.