Discrimination of phthalate species using a simple phage-based colorimetric sensor in conjunction with hierarchical support vector machine.

Here, the analytical potential of an M13 bacteriophage-based color sensor for discrimination of 4 phthalates with similar molecular structures (bis-(2-ethylhexyl)-phthalate (BEHP), dibutyl phthalate (DBP), diethyl phthalate (DEP), and benzyl-butyl-phthalate (BBP)) was investigated. The pattern and magnitude of the RGB color changes were different depending on the functional groups present in the phthalate structures. For example, BEHP possessing a long alkyl chain resulted in a minute color change, while the variation of color was substantially large when BBP containing an additional benzene ring was measured. Since a tryptophan-histidine-tryptophan residue possessing indole and imidazole was present on the self-assembled phages, the π-π interaction of benzene in BBP with the sensor surface produced a considerably greater color change. To evaluate the multi-modally varying color signals due to diverse interactions of the phthalates with the sensor and to discriminate them, support vector machine (SVM), which can construct a boundary hyperplane among complexly scattered sample groups, was used. In addition, hierarchical SVM (H-SVM) was adopted to deal with multi-class discrimination. The use of H-SVM improved the discrimination accuracy up to 90.1%, compared to 87.1% using SVM. The demonstrated color sensor is versatile and can be potentially adopted as an on-site screening tool. Strategies to improve the accuracy further for real applications are also discussed.