Computational methods for the analysis of chemical sensor array data from volatile analytes

Prospects of second generation artificial intelligence tools in calibration of chemical sensors

Multivariate data driven calibration models with neural networks (NNs) are developed for binary (Cu++ and Ca++) and quaternary (K+, Ca++, NO3- and Cl-) ion-selective electrode (ISE) data. The response profiles of ISEs with concentrations are non-linear and sub-Nernstian. This task represents function approximation of multi-variate, multi-response, correlated, non-linear data with unknown noise structure i.e. multi-component calibration/prediction in chemometric parlance. Radial distribution function (RBF) and Fuzzy-ARTMAP-NN models implemented in the software packages, TRAJAN and Professional II, are employed for the calibration. The optimum NN models reported are based on residuals in concentration space. Being a data driven information technology, NN does not require a model, prior- or posterior- distribution of data or noise structure. Missing information, spikes or newer trends in different concentration ranges can be modeled through novelty detection. Two simulated data sets generated from mathematical functions are modeled as a function of number of data points and network parameters like number of neurons and nearest neighbors. The success of RBF and Fuzzy-ARTMAP-NNs to develop adequate calibration models for experimental data and function approximation models for more complex simulated data sets ensures AI2 (artificial intelligence, 2nd generation) as a promising technology in quantitation.

Laboratory and field evaluation of a SAW microsensor array for measuring perchloroethylene in breath

This article describes the laboratory and field performance evaluation of a small prototype instrument employing an array of six polymer-coated surface acoustic wave (SAW) sensors and a thermal desorption preconcentration unit for rapid analysis of perchloroethylene in breath. Laboratory calibrations were performed using breath samples spiked with perchloroethylene to prepare calibration standards spanning a concentration range of 0.1-10 ppm. A sample volume of 250 mL was preconcentrated on 40 mg of Tenax GR at a flow rate of 100 mL/min, followed by a dry air purge and thermal desorption at a temperature of 200 degrees C. The resulting pulse of vapor was passed over the sensor array at a flow rate of 20 mL/min and sensor responses were recorded and displayed using a laptop computer. The total time per analysis was 4.5 min. SAW sensor responses were linear, and the instrument's limit of detection was estimated to be 50 ppb based on the criterion that four of the six sensors show a detectable response. Field performance was evaluated at a commercial dry-cleaning operation by comparing prototype instrument results for breath samples with those of a portable gas chromatograph (NIOSH 3704). Four breath samples were collected from a single subject over the course of the workday and analyzed using the portable gas chromatograph (GC) and SAW instruments. An additional seven spiked breath samples were prepared and analyzed so that a broader range of perchloroethylene concentrations could be examined. Linear regression analysis showed excellent agreement between prototype instrument and portable GC breath sample results with a correlation coefficient of 0.99 and a slope of 1.04. The average error for the prototype instrument over a perchloroethylene breath concentration range of 0.9-7.2 ppm was 2.6% relative to the portable GC. These results demonstrate the field capabilities of SAW microsensor arrays for rapid analysis of organic vapors in breath.

Flavor analysis in a pharmaceutical oral solution formulation using an electronic-nose

Flavors are commonly used in pharmaceutical oral solutions and oral suspensions to mask drug bitterness and to make the formulation more palatable. Flavor analysis during product development is typically performed by human organoleptic analysis, which is often expensive and less objective. A novel approach using a metal oxide sensor-based instrument (electronic-nose) for headspace analysis was explored to replace human sensory perception for consistent qualitative and quantitative analysis of flavors in a pharmaceutical formulation. The use of the electronic-nose technique to qualitatively distinguish among six common flavoring agents (raspberry, red berry, strawberry, pineapple, orange, and cherry) in placebo formulations was demonstrated. The instrument was also employed to identify unknown flavors in drug formulation placebos. Raspberry flavor samples from different lots made by the same manufacturer, as well as freshly prepared and aged samples, were also distinguished by electronic-nose. Therefore, the instrument can potentially be used for identity testing of different flavor raw materials and the flavored solution formulations. The electronic-nose was also employed successfully for quantitative analysis of flavors in an oral solution formulation. The quantitative method might be used to assay the flavor concentration during release testing of the oral solution formulation or to monitor flavor shelf-life in the marketed container. It can also be implemented for packaging selection for the formulation in order to ensure the flavor shelf-life. Chemometric methodologies including principal component analysis (PCA), discriminant factorial analysis (DFA), and partial least squares (PLS), were used for data processing and identification.

Different approaches to multivariate calibration of nonlinear sensor data

In this study, different approaches to the multivariate calibration of the vapors of two refrigerants are reported. As the relationships between the time-resolved sensor signals and the concentrations of the analytes are nonlinear, the widely used partial least-squares regression (PLS) fails. Therefore, different methods are used, which are known to be able to deal with nonlinearities present in data. First, the Box-Cox transformation, which transforms the dependent variables nonlinearly, was applied. The second approach, the implicit nonlinear PLS regression, tries to account for nonlinearities by introducing squared terms of the independent variables to the original independent variables. The third approach, quadratic PLS (QPLS), uses a nonlinear quadratic inner relationship for the model instead of a linear relationship such as PLS. Tree algorithms are also used, which split a nonlinear problem into smaller subproblems, which are modeled using linear methods or discrete values. Finally, neural networks are applied, which are able to model any relationship. Different special implementations, like genetic algorithms with neural networks and growing neural networks, are also used to prevent an overfitting. Among the fast and simpler algorithms, QPLS shows good results. Different implementations of neural networks show excellent results. Among the different implementations, the most sophisticated and computing-intensive algorithms (growing neural networks) show the best results. Thus, the optimal method for the data set presented is a compromise between quality of calibration and complexity of the algorithm.

Detection and Classification of Organophosphate Nerve Agent Simulants Using Support Vector Machines with Multiarray Sensors

The need for rapid and accurate detection systems is expanding and the utilization of cross-reactive sensor arrays to detect chemical warfare agents in conjunction with novel computational techniques may prove to be a potential solution to this challenge. We have investigated the detection, prediction, and classification of various organophosphate (OP) nerve agent simulants using sensor arrays with a novel learning scheme known as support vector machines (SVMs). The OPs tested include parathion, malathion, dichlorvos, trichlorfon, paraoxon, and diazinon. A new data reduction software program was written in MATLAB V. 6.1 to extract steady-state and kinetic data from the sensor arrays. The program also creates training sets by mixing and randomly sorting any combination of data categories into both positive and negative cases. The resulting signals were fed into SVM software for "pairwise" and "one" vs all classification. Experimental results for this new paradigm show a significant increase in classification accuracy when compared to artificial neural networks (ANNs). Three kernels, the S2000, the polynomial, and the Gaussian radial basis function (RBF), were tested and compared to the ANN. The following measures of performance were considered in the pairwise classification: receiver operating curve (ROC) Az indices, specificities, and positive predictive values (PPVs). The ROC Az) values, specifities, and PPVs increases ranged from 5% to 25%, 108% to 204%, and 13% to 54%, respectively, in all OP pairs studied when compared to the ANN baseline. Dichlorvos, trichlorfon, and paraoxon were perfectly predicted. Positive prediction for malathion was 95%.

Rapid identification of viable Escherichia coli subspecies with an electrochemical screen-printed biosensor array

Rapid identification of Escherichia coli strains is an important diagnostic goal in applied medicine as well as the environmental and food sciences. This paper reports an electrochemical, screen-printed biosensor array, where selective recognition is accomplished using lectins that recognize and bind to cell-surface lipopolysaccharides and coulometric transduction exploits non-native external oxidants to monitor respiratory cycle activity in lectin-bound cells. Ten different lectins were separately immobilized onto porous membranes that feature activated surfaces (ImmunodyneABC). Modified membranes were exposed to untreated E. coli cultures for 30 min, rinsed, and layered over the individual screen-printed carbon electrodes of the sensor array. The membranes were were incubated 5 min in a reagent solution that contained the oxidants menadione and ferricyanide as well as the respiratory substrates succinate and formate. Electrochemical oxidation of ferrocyanide for 2 min provided chronocoulometric data related to the quantities of bound cells. These screen-printed sensor arrays were used in conjunction with factor analysis for the rapid identification of four E. coli subspecies (E. coli B, E. coli Neotype, E. coli JM105 and E. coli HB101). Systematic examination of lectin-binding patterns showed that these four E. coli subspecies are readily distinguished using only five essential lectins.

Differential receptors create patterns diagnostic for ATP and GTP

Herein we report the combination of a library of resin-bound sensors along with a multicomponent sensor array. This novel combinatorial array sensor system shows selectivity for nucleotide phosphates in solution. The design of the anchored receptor includes a 1,3,5-trisubstituted-2,4,6-triethylbenzene scaffold coupled with peptide libraries. Each chemosensor is placed into a micromachined cavity within a silicon wafer, and the optical changes observed by a charged-coupled device result in near-real-time digital analysis of solutions. A colorimetric displacement assay was performed, and time-dependent imaging studies of the selected sensing ensembles result in a differential responses upon addition of adenosine 5'-triphosphate (ATP), adenosine 5'-monophosphate (AMP), or guanosine 5'-triphosphate (GTP). An advantage to this approach is that it creates an array of sensors that gives a fingerprint response for each analyte. Principal component analysis indicates that the library of chemosensors can differentiate between ATP, GTP, and AMP. On the basis of factor loading values, individual sensors from the library were sequenced to elucidate their chemical composition.

Imaging optical sensor arrays

Imaging optical fibres have been etched to prepare microwell arrays. These microwells have been loaded with sensing materials such as bead-based sensors and living cells to create high-density sensor arrays. The extremely small sizes and volumes of the wells enable high sensitivity and high information content sensing capabilities.

"Electronic nose" detects major histocompatibility complex-dependent prerenal and postrenal odor components

Mice prefer to mate with individuals expressing different MHC genes from their own. Volatile components presenting MHC-dependent odor types are present in urine and can be detected by mice, as shown by extensive behavioral studies. Similar odor types are suspected to influence human behavior as well. Although a recent report indicates that MHC expression influences the ratio of volatile compounds such as phenylacetic acid, so far no other means than studying the behavior of mice or rats has been available to assess odor types. Here, we report the ability of a gas sensor array (referred to as "electronic nose") to detect MHC-dependent odor types. The electronic nose consists of an array of chemophysical detectors, in our case quartz crystal microbalances and semiconducting metal-oxide sensors that change frequency or conductivity upon binding of very small numbers of individual molecules present in the gas phase of odorous fluids. The pattern of changes is characteristic for a particular smell. Our electronic nose distinguishes the urine odor types of MHC congenic mouse strains, MHC class I mutant mice, and HLA-A2 transgenic mice. In addition, MHC-dependent odor types can be detected in serum. The device also clearly differentiates between individual odor types of human sera from HLA homozygous individuals; however, HLA expression seems to have only a secondary influence. Thus, odor-type research can now be carried out with an objective and fast through-put system independent of behavioral studies.