We propose to establish reliable technologies for monitoring quality traits and detect and quantify potential microbial/chemical hazards using vibrational spectroscopic techniques, including FT-IR and Raman spectroscopy. The objectives of our proposed research focus on developing:Validated prediction models generated with customized vibrational spectroscopy technology to the specifications required by the industry for multiple quality traits.Develop spectral libraries for threat materials and fingerprinting microbial and chemical hazards in food matrices.Our proposed research will focus on mature spectroscopic sensor technologies, to enable pathogen identification with analytical precision equivalent to bench-top instruments found in the laboratory. We will also use our find ways to science and engineering expertise to improve or expand upon these existing technologies to meet challenges specific to food monitoring (e.g., front-end sampling, purpose-built sensor systems and detection algorithms, spectral libraries).Successful development of such real-time and field-based spectroscopic systems will provide significant advantages for food safety monitoring. These technologies can enable greater control of the high volume raw material stream, addressing both safety and brand equity. Implementation by the industry and regulatory agencies of rapid testing procedures based on these technologies would help to streamline food safety and quality assurance and will prevent the growing danger to the health of consumers from adulterated or substituted products as evidenced by the melamine incident. Development of spectral libraries to fingerprint high-risk fresh food crops (i.e. spinach, alfalfa sprouts, and tomatoes) and ingredients (i.e. powdered milk, corn meal, cocoa & peanut butter, fruit juices) would provide the food industry unique detection capabilities to prevent fraud and ensure consumer product safety. Testing done as close to the original source would permit detecting adulteration before an ingredient has been diluted or combined with other ingredients. Vibrational sensing technologies would permit interrogation of produce wash water for chemicals and pathogens to evaluate exposure.
The experimental design will include establishing a comprehensive reference spectroscopic profile for materials/ingredients, model design and analysis of samples using FT-IR, and Raman spectroscopy. We will identify an appropriate sample set to establish a reference spectral range for materials from identified manufacturers. The pure samples will be combined to form a reference spectral range. These samples will be analyzed using FT-IR and Raman spectroscopy, and the spectra will be evaluated for highly specific chemical signatures to characterize particular component make-up.Monitoring quality traits: We will work with economically relevant products (ie. tomato juice, vegetable oils, cocoa butter, among others). Spectra of samples will be collected by using different infrared accessories (transmittance, ATR, diffuse reflectance). Important quality traits will be determined by official reference methods. All spectral data will be analyzed by chemometrics using Partial Least Squares Regression. Multivariate analysis will model relationships between the infrared spectra and quality parameters and its predictive accuracy will be examined with an independent validation set and the accuracy of the models determined.Method sensitivity and quantification microbial/chemical hazards: We will develop robust and sensitive protocols that will allow the rapid quantitation of chemical contaminants and pathogens by Infrared/Raman spectroscopy to yield specific signal of marker compounds. For microbial detection, we will use immunomagnetic separation for target pathogens in high-risk produce contaminated at different levels (10, 102, 103, and 104 CFU/g) to determine level of detection. The magnetic affinity-captured microorganism are applied onto an array slide and analyzed by reflectance Infrared/Raman spectroscopy. For each target pathogen, triplicate samples will be prepared at each level of contamination to evaluate reproducibility. The experiment will be conducted on different days (at least 4 days) to evaluate the robustness of the spectral data. The vibrational spectroscopy technique will provide a tool to current rate-limiting and time-consuming methods to effectively evaluate the safety of produce/ingredients. Furthermore, we expect that the low sensitivity and high accuracy and reproducibility of the IRMS method will provide levels of detection <10 CFU/g. Chemical Contaminants: Uncontaminated high-risk samples (ie. powdered milk, cocoa & peanut butter, fruit juices) will be spiked with different adulterant-like chemicals. It is important that the samples be homogenous so the samples will be thoroughly mixed before analyzing. The model design will be based on two factors: adulterant and concentration. The spiked samples will be analyzed using FT-IR and Raman spectroscopy. Spectral analysis will be carried out by supervised chemometric methods for sample classification (SIMCA) and quantitative (PLSR) analysis. SIMCA consists of assigning training data sets to classes and then a principal component model is created for each class with different confidence regions. Probability clouds (95% CI) are built around the clusters based on PCA scores, allowing SIMCA to be used as a predictive modeling system. This step also involves determining the lower limits of adulterant concentration that each technology can detect. For each chemical entity, triplicate samples will be prepared at each level of contamination. Quantitative models based on infrared spectra will be generated by Partial Least Squares regression (PLSR) using the spiked levels confirmed by using a reference method. Using published data on fundamental vibrations of specific functional groups and available standards will correlate the IR spectral signals. Classification and regression models will be used to generate prediction models and the accuracy and ability of these models will be examined with an independent test set representative of the classes modeled with the training set. Blind samples (the researcher will not have access to its identity before prediction) will be included to test the ability of the models to predict the identity and levels of a potential chemical contaminant.Deployability of protocols/models: We will consider existing technologies such as newly developed portable optical systems for chemical identification which bring the analytical precision of spectroscopy to field applications with spectral resolution equivalent to bench-top instruments. We will also find ways to improve or expand upon existing technologies (front-end sampling, purpose-built sensor systems and detection algorithms, spectral libraries) using our expertise to make the technologies more applicable.