Application of TargetView software within the food industry - the identi cation of pyrazine compounds in potato crisps

Transcription

1 Application Note: ANTV10 Application of TargetView software within the food industry - the identi cation of pyrazine compounds in potato crisps Abstract In order to regulate product safety and quality within the food industry, methods for measuring speci c volatile organic compounds (VOCs) in a product are implemented. Such analysis is typically performed by thermal desorption (TD) with GC/MS, inevitably resulting in high sample load. In a manufacturing environment, interpretation of the complex data for the identi cation of trace-level target compounds is time consuming and often inaccurate. To address this, new software has been speci cally developed to automate this process, therefore saving time and improving accuracy. Introduction For safety and quality control purposes, there is growing demand within industry to detect and identify speci c volatile organic compounds (VOCs), e.g. in products or the surrounding environment. The food industry is a good example of this with the advent of various regulations involving product safety and quality, typically in relation to avour composition, taint and contamination. Standard VOC analyses of food samples by TD-GC/MS can yield complex chromatograms in which some compounds, particularly those at trace level, are hard to detect. Interpretation of this data using conventional library search methods is very time consuming and unlikely to generate accurate results, particularly for laboratories/ manufacturers with limited MS expertise. To address this problem, TargetView software (ALMSCO International Ltd., UK) has been developed to automate interpretation, therefore simplifying and speeding up analysis. After importing the GC/MS data le for the sample, the software interrogates the total ion chromatogram (TIC) for compounds of interest (e.g. regulatory compounds) and generates a simple post analysis report of those identi ed. TargetView processes GC/MS data using de ned target libraries of compounds as a point of reference. The number of compounds within these libraries may extend from an individual substance up to several hundred. Within complex GC/MS TIC pro les there may be signi cant compound coelution which may also include matrix effects. In an attempt to alleviate these problems, TargetView adopts a chemometric approach to data analysis. The software initially minimises background noise by dynamic background compensation (DBC). This is then followed by spectral deconvolution to identify individual components within the data, and nally, principle component analysis (PCA) highlights target compounds found in the sample by comparing the deconvolved compound spectra to those of the target library. To illustrate these processes and demonstrate the data-mining ef ciency of the software via its application within the food industry, a sample of potato crisps was analysed for the presence of speci c pyrazines. Pyrazines are VOCs that are responsible for avouring in a variety of foods. The extraction of chemicals from the crisps using a Micro- Chamber/Thermal Extractor (µ-cte ; Markes International Ltd., UK) leads to sensitive, and therefore comprehensive, VOC analysis by TD-GC/MS. Method A crushed sample (~5 g) of potato crisps was placed into one of the four chambers of the µ-cte unit (four- and six-chamber µ-cte units are available). The chambers were heated to 40 C and a purge gas of helium was applied at a ow rate of 40 ml/min for 25 minutes to allow equilibration. Air exhausting from the chamber was then collected for 10 minutes onto a 3.5 x 6.4 mm OD thermal desorption (TD) tube (Markes) containing Tenax TA as the retaining sorbent. Trapped emissions were thermally desorbed on UNITY 2 (Markes) and analysed by GC/TOF-MS. Analytical conditions TD Instrument: UNITY 2 Desorb temp: 300 C for 5 min Trap low: 0 C Trap high: 320 C Trap: General purpose trap (Tenax TA) Split: Single split 20:1

2 ANTV10 April GC Oven: Run Time: Column: Initial ow: 40 C for 2 min, then 20 C/min up to 240 C for 2 min 14 min HP-INNOWax Polyethylene Glyco 30 m x 250 µm x 0.25 µm 1.5 ml/min MS Instrument: BenchTOF-dx Transfer line temp:250 C Ion source temp: 200 C Results and discussion Report generation The GC/MS data le is imported into TargetView and the TIC processed. This results in an Excel-compatible text report, listing target compounds found within the sample. Several pyrazines were identi ed in the crisps by the software. However here, three speci c compounds (dimethyl pyrazine, ethyl methyl pyrazine and ethyl dimethyl pyrazine) were selected for reference as their concentration was very low (Figure 1). The information in the report can be sorted by compound name, retention Figure 2: GC/MS data for crisp sample showing the effect of DBC (original data shown in black, DBC data shown in blue) time, area, etc. This useful report is automatically generated and easily interpreted by a non-expert user. Implementation of DBC The initial dynamic background compensation (DBC) process is fundamental to the pro ciency of TargetView. The purpose of DBC is to suppress background ions that are not contributing towards TIC peaks, i.e. interference or noise from column bleed, air/water offset, etc. This is the rst algorithm applied to the data, and the result is a signi cantly improved baseline with enhanced compound spectral purity. This allows TargetView to process real peak data without the inclusion of interfering ions. As TargetView preserves the original data le (creating a second le for processing), the DBC-processed data can be compared to the original to observe the improvement. Figure 2 shows the effect of DBC on the TIC for the crisp sample; the black trace represents original data, and the blue represents the DBC data. The enhancement of the TIC pro le as a result of DBC-processing is clear. Figure 2 also indicates the high sample load from the TD system into the GC/MS. This is re ected in the abundance scale, which ultimately extends to 1+e8. The large concentration of some components in the sample makes identi cation of trace level compounds much more challenging; especially where co-elution is present. Deconvolution and PCA Figure 1. Example post-run report for three pyrazines found in crisp sample Subsequent to DBC, spectral deconvolution is applied to the TIC. This provides enhanced compound spectra for the principal component analysis (PCA) step. Where coelution occurs, spectral deconvolution processes the data for every scan/scan-set across the peak and assigns ions to the appropriate component. Conventional deconvolution techniques can be adversely affected by matrix effects from the sample; however the initial DBC stage of TargetView has

3 ANTV10 April alleviated this. Prior to deconvolution, attempting to identify compounds by library-searching spectra obtained from overlapping peaks would be likely to result in incorrect/unreliable matches. PCA is the nal stage of the automated TargetView process. Advanced chemometrics are applied to the deconvolved spectra to highlight characteristic ion fragmentation patterns, which are then compared to the compound spectra of the target library to determine matches. A match coef cient is calculated for each target analyte ranging from 0 1. The higher the coef cient, the more reliable the result; however, compounds with a match coef cient below a user selectable value (e.g. 0.8) can be excluded from the nal report to enhance con dence in the results. Graphical user interface Figure 4: Trace-level identi cation of three pyrazine compounds (indicated by arrows) Figure 3 shows the TargetView graphical user interface (GUI), comprising an upper window showing the DBC TIC, and a lower window showing the match coef cient plot for a speci c target compound with respect to retention time. In this example, dimethyl pyrazine has been selected from the drop-down list of target compounds. The presence of a bar in the lower window indicates that it has been identi ed and the co-ef cient (0.963) indicates a very reliable match. In order to visualise compounds found in the sample, the upper TIC displays an overlaid histogram plot (HPlot) in which compounds are represented by red bars. The HPlot can be con gured to show target hits only or all components. From this overlay, the degree of coelution can be seen. In addition, the bar height provides a representation of peak area. This is based on the summation of user-selected certain ions from within the deconvolved spectra. The area highlighted in gure 3 is enlarged in gure 4 to display the HPlot overlay for the target compounds. The location of the three speci c pyrazines can be seen; namely, dimethyl pyrazine, ethyl methyl pyrazine and ethyl dimethyl pyrazine. Note also the additional pyrazine compounds present in the crisps and identi ed by the software, these are listed in the upper left hand window of the GUI. The HPlot bar heights (relative to TIC peak height) indicate that these compounds are at very low concentration and submerged in an extremely complex sample matrix. By con guring the display to show the HPlot overlay for all components, i.e. both targets and unknowns, the degree of coelution can be seen. Figure 5 shows the location of the lowest concentration target compound (ethyl methyl pyrazine, labelled T ), which is co-eluting with three other unknown components. Figure 3: TargetView GUI (upper window shows DBC TIC, lower window shows match co-ef cient plot for dimethyl pyrazine)

4 ANTV10 April Identifying unknowns TargetView is primarily used to detect target compounds; however it can also be useful for the identi cation of unknowns within the sample. Figure 5: TargetView HPlot overlay con gured to display all components (red bars). Note the co-elution of ethyl methyl pyrazine (T) with three other compounds in the same peak. Despite the high coleution and low concentration, the match co-ef cient for this compound is very high (0.96). By using conventional search methods on the raw GC/MS data, such a compound would be impossible to reliably identify. Match con dence and NIST con rmation Within TargetView, the deconvolved (apex) spectrum for a compound can be directly compared to the library entry spectrum. Figure 6 shows an example of this for the compound dimethyl pyrazine. The sample spectrum (top window) compares very well with the library spectrum (bottom window), producing the high match value (0.963). To cross-check the validity of the result from TargetView, the sample spectrum can be searched against a commercial database, e.g. NIST. Figure 7 represents the NIST search results for the dimethyl pyrazine spectrum, which indicates a high probability, rst-hit for this compound. The high match co-ef cient derived from TargetView and the additional NIST con rmation demonstrate the datamining ef ciency of the software. The identi cation of trace level compounds that are co-eluting with much higher concentration analytes is testament to the multistep approach (i.e. DBC, deconvolution and PCA) of TargetView. Figure 8 shows the spectrum for an unknown compound coeluting with the target compound diethyl pyrazine. The relevant area of the TIC is displayed inset, which shows multiple compounds eluting within the single peak. This unknown is represented by the symbol U and diethyl pyrazine represented by the symbol T. The relative heights of the HPlot bars show a signi cantly higher concentration for the co-eluting compound U over the target compound T. By searching the spectrum of an unknown within the NIST database it may be possible to identify the substance. Figure 9 represents the NIST analysis of the unknown component U spectrum. A strong match was found meaning the compound was identi ed as dimethyl styrene with high con dence. Figure 6: Comparison of sample and target spectra for dimethyl pyrazine Figure 7: Cross-searching deconvolved sample spectrum for dimethyl pyrazine against NIST database con rms dimethyl pyrazine

5 ANTV10 April Figure 8: Spectrum of unknown compound U1. Inset: HPlot overlay of target (T) and unknown compound (U). Conclusions The analysis of food products using the µ-cte and TD-GC/TOF-MS generates a high load sample, resulting in a complex GC/MS TIC pro le. The identi cation of trace-level targets within this convoluted matrix would not be possible using conventional library searching techniques. In this application, the spectral deconvolution and advanced chemometrics of TargetView have resulted in the successful identi cation of trace-level pyrazines. This is a clear demonstration of the intensive datamining capabilities afforded by the software. In addition, unknown compounds may also be identi ed if necessary. With the advent of product quality and safety regulations requiring the monitoring of speci c VOCs, TargetView is an ef cient software tool for users with limited MS expertise. This translates into increased productivity, saving time and ultimately cost in the VOC analysis of a product. Figure 9. NIST identi cation of unknown component (U) co-eluting with target TargetView has many potential applications including food, avour and fragrance analysis, materials emissions testing, air toxics detection, etc.