GC-MS or LC-MS (or both) database of alcoholic beverages?

[bisnettrj2]

I got to thinking the other day after watching an episode of the American television program "Bones", wherein part of the plot involved a character procuring vintage empty wine bottles and subsequently refilling those bottles with more recent vintages of the same wine, recorking the bottle, and selling it as the vintage product.

Is there currently any sort of database for cataloguing the various vintages of different wines, high-end spirits, or craft beers by means of GC-MS or LC-MS, to verify the identity of the initial product 10,20, or 50 years from now? How much would the cellaring or aging process change the inherent profile of the initial product? Would any such database even be possible with the change in character of the initial product due to the aging process? Are there components of each product (wine, liquor, beer) that would not change over time, and might act as a "fingerprint' to identify a specific vintage, in order to prevent fraudulent trade in vintage beverages? If anyone has any insight, I would be interested in reading your thoughts on the issue.

[rhaefe]

Only thing that comes to mind right now is the quantification of tritium to determine the age of wine. No idea if this is still done.

http://www.pnas.org/content/39/4/245.full.pdf

[bisnettrj2]

I just came across this study a few weeks ago...

http://pubs.acs.org/doi/abs/10.1021/jf102632g

Abstract
The aim of this study was to asses the applicability of LC-MS profiling to authenticate a selected Trappist beer as part of a program on traceability funded by the European Commission. A total of 232 beers were fingerprinted and classified through multivariate data analysis. The selected beer was clearly distinguished from beers of different brands, while only 3 samples (3.5% of the test set) were wrongly classified when compared with other types of beer of the same Trappist brewery. The fingerprints were further analyzed to extract the most discriminating variables, which proved to be sufficient for classification, even using a simplified unsupervised model. This reduced fingerprint allowed us to study the influence of batch-to-batch variability on the classification model. Our results can easily be applied to different matrices and they confirmed the effectiveness of LC-MS profiling in combination with multivariate data analysis for the characterization of food products.

[lmh]

further to the above (!) I vaguely remember going to a conference session where a manufacturer presented some multivariate work on beers (I think largely because the samples were readily available, and disposal of excess sample was cheerfully staff-friendly).

As I remember, there was a curious outcome: it was easier to distinguish two very similar beers from different brewers than it was to distinguish two very different products from the same brewery. Clearly analytical instrumentation and humans have a different approach to beer.

I don't know if they ever published the work, and can't even remember the company now (sorry).

[R13]

Short answer: GC-MS and LC-MS are not suitable for age determination. Even the determination of "identity" is a problem - wines are too complex, vodka - too simple. In all the drinks (and many foods) the most important differences are in "small print" - minor components which might be organic but also inorganic compounds.

On the other hand HPLC-MS spectra are dependant on the model/producer/ionisation mode so the library can be made only for the particular instrument - which means all the reference samples must be re-measured in each lab and when new instrument arrives. You can imagine cost of this....

Geograpical location of production can be established using special deuterium NMR-SNIF techiques and correlation analysis - and it is being introduced in forgery investigation, but price is still to high for routine use. Using it they can say if the wine (or even mineral water) comes from the particular region/loacation.

I'm not specialist, but guess that (commonly available) radioactive decay based measurements have too big of uncertainty to distinguish between say 20, 18 and 10 years old samples.