Automated Component Detection and Matrix Analysis of Time Resolved MS data – Analysis of Wine Volatiles by GC/MS

Automated Component Detection and Matrix Analysis of Time Resolved MS data – Analysis of Wine Volatiles by GC/MS

J.H. Moncur1, S.J. Campbell1 and A. Hall2

1SpectralWorks Ltd, The Heath Business & Technical Park, Runcorn, UK; 2Hall Analytical Ltd, Floats Road, Manchester, UK

First Published: BMSS 2007

Abstract

The process of determining components by hand, particularly minor components in unknown samples, from GCMS and LCMS data is tedious, time consuming and may be subjective from day to day and from analyst to analyst. In many of the ‘Omics’ applications the analyst may not yet know what the key variables are, therefore consistent and in depth data mining is essential.

The ability to review time resolved MS data, such as GC or LC across a number of similar samples in an objective and quantitative manner was presented. Samples of wine were analyzed by GC/MS to observe how the pattern of volatile components changed with time following initial opening and pouring. Chromatographic component determination was carried out using a proprietary software algorithm whereby a component spectral library was created to allow a qualitative matrix analysis between the samples. The use of an internal standard permits quantitative data to be presented with respect to changes in, rather than absolute, concentrations of components. This allowed the tracking of ‘Unidentified’ components across the matrix.

Introduction

Wine is complex. The are many factors during the manufacture, storage and maturation of wine which affect the taste and flavour. How it is served, whether decanted or not and at what temperature etc are all considered to influence the appreciation of the wine which in itself can be particularly subjective. However, all wine begins to change once it has been opened not that change is necessarily bad. A young vintage port wine may need to be decanted a day ahead of time just to get it into shape for drinking, while a delicate Spanish Moscatel might be considered to have started to decline in quality after just a few hours. For many wines, there will be more alcohol flavour the longer it sits, others will lose all flavour and taste like water and yet others will become more acidic. It depends on the wine.

The goal of this analysis was to show that a component detection algorithm and matrix analysis can provide an objective assessment of the change in wine volatiles. To demonstrate this, the headspace volatiles from an Italian wine were measured using solid phase microextraction (SPME) GC/MS over an extended period of time. The wine used was a 2005 Montepulciano d’Abruzzo. It is made primarily from the Montepulciano grape that makes wines of medium body, rich colour and nice tannin structure (Cellarnotes.net).

Materials and Method

The sample bottle of wine was transferred to a 1L Duran type bottle with a magnetic stirrer bar and spiked with 10 µg of deuterated benzene. The temperature was maintained at 23ºC and the stirring rate was set at 75 rpm. The SPME fibre was set at 3 cm above the wine and exposed for 30 minutes. A method and needle blank was run using de-ionised water. The SPME fibre was desorbed in the GC injection port and the volatile components were cold-trapped using liquid nitrogen on a 0.5 metre length of the analytical column.

SPME Fibre type:Carboxen/PDMS 75 µm (Supelco # 57318)
GC/MS:Agilent 6890 MSD Splitless injection
Column::60 m x 0.25 mm (0.25 µm) DB-5 MS
Flow Rate:1mL/min constant flow
Source Temperature:230 ºC
Full Scan Acquisition:m/z 20 – 500 at 3 scans/s
Injector Temperature:300 ºC
GC Temperature Programme:45 ºC for 1 min to 100 ºC at 5 ºC/min to 300 ºC at 20 ºC/min for 3 min

Nine SPME samples were taken over at the elapsed times shown in table 1. Data processing and analysis was carried out using AnalyzerPro™ with MatrixAnalyzer™.

Table 1. SPME Sample times (elapsed)

Sample123456789
Sample04080120160200240280320

The components detected in the first SPME sample were library searched for identification and naming purposes and used to create a target library. A light filtering was applied to remove a number of the minor components at this time. A matrix analysis of these components was performed across all of the SPME samples. An analysis of the change in ratio of the components to the deuterated benzene was included.

Results

Figure 1 shows the TIC from the fist (upper) and the last (lower) headspace sample.

Due to the overloaded nature of the ethanol component, analysis was restricted to components which eluted after approximately 4.5 minutes. The later part of the run after 21 minutes was also excluded as the volatile components had already eluted.

Table 2 shows the MatrixAnalyzer report for the nine SPME samples. For each sample the absolute peak areas and the peak areas relative to the deuterated benzene (lighter text) are shown. The final set of three columns show the average response, the standard deviation and the co-efficient of variation of the responses.

Conclusions

A number of subtle differences in both the absolute levels of components as well as the relative ratios of some of the components can be seen from Figure 1.

Figure 1. TIC of first and ninth SPME samples.

Table 2. Matrix Analysis of SPME Samples

Other than a single unidentified component (number 28) the deuterated benzene had the largest absolute coefficient of variation of 152%, over twice that of the next largest variation (component 53). This shows that the deuterated spike did not reflect the behaviour of any of the native volatile components. Subsequent analysis of the relative peak area responses was not carried out as it was felt that this would only reflect the variation of the deuterated spike itself.

The remaining components had a variation of between approximately 5 and 60% (mean = 16 % n = 52). The report from MatrixAnalyzer allows for quick ranking to determine the components showing the greatest change.

This work shows the proof of concept of being able to set up an objective matrix analysis using the capabilities of AnalyzerPro to detect clean components without prior knowledge of the likely constituents. Going forward, the same approach could be applied to a more detailed component analysis of the wine. An improved sampling technique which allows for the use of an ‘injection’ standard would assist with providing data on the relative changes in the wine. Additional work in determining the identity and source of the component 28 will be investigated.