A novel single-particle multiple-signal sensor array combined with multidimensional data mining for the detection of tricarboxylic acid cycle metabolites and discrimination of cells

Tricarboxylic acid (TCA) metabolites in cancer cells show a marked difference from those in normal cells. Herein, we report a single-particle multiple-signal lanthanide/europium-based metal–organic framework (Tb/Eu MOF) sensor array for the detection of TCA metabolites and discrimination of cancer cells. In the presence of TCA metabolite, 6 characteristic peaks of Tb/Eu MOF showed dramatic changes due to host–guest interactions, allowing sensor array-based qualitative and quantitative detection to be performed. In the qualitative detection ability test, 18 TCA metabolites at 4 concentrations (50 μM, 100 μM, 200 μM, 300 μM) were accurately discriminated by the sensor array via linear discriminant analysis (LDA). Significantly, these 4 concentrations include the clinical detection criteria for most TCA metabolites. In the quantitative detection ability test, a good linear relationship between Euclidean distances and the concentrations of L-valine (Val) could be obtained in the range of 50 to 500 μM (R(2) = 0.9755). On this basis, the provided method was successfully applied for the classification of 2 normal cells and 5 cancer cells via principal components analysis (PCA), LDA and a radial basis function neural network (RBFN). What's more, by verifying the weight coefficient of each point, detection and discrimination results are proved as a trustworthy balanced evaluation of multiple factors. Depending on precise data processing, the experimental operation was simplified on the premise of ensuring accuracy, so our method is a meaningful exploration for array design. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s00216-023-04736-1.