Chemosensor devices for detection and quantification of analytes using multivariate analysis and machine learning based approaches

Abstract

Chemosensors are molecular systems that interact with particular target species to generate an analytically useful signal. For instance, the optical chemosensors interact with analytes resulting a change in optical properties (absorption, fluorescence) which is detected using traditional instruments such as UV-Vis absorption and fluorescence spectrometers. Although traditional instruments provide highly reliable and efficient detection of analytes however, the lack of portability, and complex sample preparation hampers the application for real time onsite detection of analytes. Therefore, the present research work is focused on the development of materials, sensing strategies, and affordable devices for onsite detection and quantification of biologically and environmentally relevant analytes. The first chapter introduces the chemosensors and the type of chemosensors based on the nature of the output signal produced on the interaction with analyte. Further, the designing of optical chemosensors has been discussed along with the various strategies and concepts employed for development of colorimetric as well as fluorometric chemosensors. The two major approaches i.e., lock-ley based single analyte sensing and olfaction/gustation inspired array-based sensing approaches has been discussed in detail. The shifting paradigm in the chemo sensing from single analyte selective sensor to cross reactive sensors-based arrays had led to development of sensor arrays for discrimination of wide range of analytes. The brief description of reported different array-based sensing system has been included. The array-based sensors generate multidimensional response on interaction with analytes, therefore role of multivariate analysis tools as well as machine learning based approached have been discussed. Finally, the summary of array-based sensing devices has been included along with working principle of each device. In the second chapter, nanomaterials-based chemosensors were developed for fluorescent and colorimetric sensing of biomolecules i.e., histidine and cysteamine in aqueous system. In the first section, Histidine sensing was achieved using carbon quantum dots (CQD) as an energy donor unit and a rhodamine 6G derivative (H1) as an energy acceptor unit. Interestingly, introduction of Fe(III) into the mixture of CQD and H1, resulted in formation of molecular ensemble and a strong fluorescence resonance energy transfer (FRET) was observed between CQD and H1. Further, the addition of histidine to an aqueous solution of ‘CQD–Fe–H1’, displaced the cationic linker Fe(III) from the ratiometric ensemble therefore resulting in turning off of the FRET. Finally, for real time application spiked urine samples were successfully analysed using ‘CQD–Fe–H1’ based sensing system. Therefore, a simple and selective system for histidine recognition in aqueous system was developed using the physical mixture of CQD and H1. Subsequently, in second section, for sensing of cysteamine a chromogenic and ratiometric sensor (H2) was developed based on azophenol derivative. The organic nanoparticles of developed receptor H2 formed chromogenic complex on interaction with Cu(II) metal ion. The colorimetric complex H2.Cu(II) further underwent a sharp colour change on interaction with cysteamine therefore provided a naked eye detection of cysteamine. Motivated by sharp colorimetric response, silica based solid state sensor strips impregnated with receptor H2.Cu(II) complex were deployed for onsite detection of cysteamine. The colorimetric response of solid-state sensors was evaluated using digital images captured by smartphone and HSV colour model. Therefore, lab-on-mobile based affordable and economic platform was developed for easy-to-use and rapid colorimetric sensing of cysteamine in aqueous as well as biological test samples. In third chapter, azophenol dye based colorimetric sensors and portable devices were developed for detection of toxic analytes (Hg(II) and CN-). In the first section, an azo-phenol based chemodosimeter (H3) was synthesised which selectively binds with Hg(II) to produce sharp colorimetric response. Further, as alternative for traditional spectrophotometer, a simple smartphone-based solution was proposed for determining concentration of Hg(II) from the change in the colour intensity of sensor impregnated solid sensing chip. The sensing chips were treated with varying concentration of Hg(II) and digital images were recorded using smartphone and analysed using RGB analysis. In order to keep all external factors constant while capturing the images of the chips, a special smartphone attachment was developed using 3D printing. Further, RGB analysis on each digital image suggested the most linear relationship between intensity of green component of colour and concentration of Hg(II) with remarkable R2 values. Therefore, it was concluded the concentration of Hg(II) could be efficiently determined using sensing chips, simple smartphone camera and an inexpensive 3D printed attachment. In the second section, highly selective azophenol-based chromogenic ligand H4 was synthesized for determination of cyanide in aqueous system. Motivated by colorimetric response of the ligand, a paper-based sensing strip was prepared capable of producing sharp colour change on interaction with cyanide. Further, a simple and economic colorimetric device was built using colour sensor (TCS3200) and Arduino microcontroller for quantification of cyanide using sensor coated paper chip. The components of the colorimeter were designed using CAD software and fabricated using 3D printer. Finally, the device was successfully calibrated using sensor coated paper chip and known concentrations of cyanide in aqueous medium. After the calibration, the applicability of device was successfully evaluated for quantification of cyanide concentration in spiked river water and food samples. Therefore, an inexpensive and portable colorimeter was successfully developed for sensing of analytes using paper-based sensing strips. The perusal of literature revealed the shifting paradigm in sensing from the single analyte selective sensors to multianalyte cross reactive arrays therefore, the fourth chapter includes array-based sensing approaches for application in food quality monitoring. In first section, a field deployable cross-reactive sensor array and a field-portable array reader has been developed for determination of biogenic amines in food samples. The sensor array was prepared using the metal complexes of single azophenol dye-based receptor H5. The response of the sensor array on interaction with multiple analytes was recorded using UV–Vis absorption spectrophotometer. The interaction of multiple sensors with multiple analytes generated complex higher dimension dataset which was evaluated using multivariate statistical techniques such as principal component analysis and linear discriminant analysis. The results of multivariate analysis suggested that single receptor-based array was capable of discriminating amongst different analytes. Therefore, motivated by outstanding discriminatory power of sensor array, the field deployable a portable reader was developed and utilized for quantification of biogenic amines in food samples. Subsequently, in second section, in continuation to previous work, a new receptor H6 was prepared utilizing azophenol based aldehyde and thiosemicarbazone derivative. The presence of thiosemicarbazone in the receptor facilitated the formation of colorimetric complexes of the receptor with the different metallic cations. The metal complexes of receptor were employed for array-based sensing for discrimination of different biogenic amines using UV–Vis absorption spectrometry. Further, the colorimetric array was optimized to reduce number of sensors using descriptive statistics. Inspired by the response of the array, the overall performance of the array-based system was evaluated using multivariate analytical tools such as principal component analysis and linear discriminant analysis. For ease of analysis, a neural network-based classifier was also developed and the model was trained and validated using k-fold cross validation routine. The results suggested that developed network was capable of discriminating the different biogenic amines with 86.1% accuracy. Therefore, an array-based sensing was performed and neural network-based classifier was successfully developed for discrimination of biogenic amines. The fifth chapter entails data modelling of effluent treatment plant and development of optical BOD sensor. In the first section, the dataset containing water quality parameters of wastewater samples collected from 4 stages of effluent treatment plant (ETP) i.e., collection tank, beverage tank, common equalization tank and final treated water tank was modelled. The dataset contained measurement of following parameters such as pH, TDS (Total dissolved solids), TSS (Total suspended solids), O&G (Oil & Grease), COD (Chemical oxygen demand), and BOD (Biochemical oxygen demand). The objective of the work was to understand the multiparametric data for better functioning and working of effluent treatment plant. Therefore, the evaluation of the complex dataset was performed and the results highlighted the strong correlation ship between different water quality parameters. Further, principal component analysis (PCA) allows to visualize the complex dataset in lower dimensions and PCA projection maps clearly shows presence of 4 major classes in the dataset which were corresponding to each stage of treatment step. Further, based on the observation, PCA based solution was proposed for fault detection and evaluation of efficiency of the treatment process. In the second section, development of an optical BOD sensor was proposed for online monitoring of the water quality in real time. The optical device works on the principle of fluorescence and the calibration of the device was performed against the BOD values of river water. The river water was periodically collected from the Sutlej River head Ropar and BOD of the sample was recorded using 3 days and 27°C temperature protocol (classical method). The collected samples were also studied using the in house developed optical BOD sensor and response of the device was recorded. Finally, the relationship between BOD values obtained using classical method and response of optical sensor was determined using machine learning based multilinear regression analysis. In the final chapter, research work was concluded along with the future prospective in field of chemosensor and chemosensor devices for onsite monitoring of analytes of interest.