Chemistry classifies coffee

Lindinger's group at the Nestle Research Center in Switzerland published a study back in March 2008 in Analytical Chemistry about analyzing coffee with mass spec. I read about this on Engadget then promptly forgot about it. I even gave a presentation on it for my Analytical class, only later to see this in my RSS bookmarks.

Basically by using PTR-MS (proton transfer reaction mass spectrometry) they did an analysis of different coffees. They emphasized that this was a data-driven study, not a chemical analysis study, because they weren't necessarily analyzing the different compounds individually. Rather, what they were doing was taking the results of the mass spec, then combining them with the 'results' of a 10-member panel of coffee experts to create a model. So they just took the intensities of the different peaks (all the compounds that had 108 m/z, 110 m/z, etc.) and compared them to the 'intensities' of the panel ratings. A rough scheme is shown below.



The panelists rated the intensity of different qualities (coffee, bitter, cocoa, roasted, woody, cereal, butter toffee, acid, citrus, winey, and flowery) of the coffee. They ran a blind study, and the panel was able to produce reproducible results, so they apparently know what they're doing. I would imagine they look something like this:



Once the model was created, they did PTR-MS on another set of coffees, had the panel do their tests, then compare how well the model was able to predict it. You can see in the graph below that they were pretty successful. You can see that there are only 8 qualities below; they decided to scrap a few qualities, but didn't really explain why. The ones that they got rid of were: winey, flowery, cereal. You can assume that they took those out because they didn't fit with the model as well, and that's probably because those qualities are made up. Nobody drinks coffee and thinks, oh that was nice and flowery. But not winey enough. Nonsense.



I'd like to see them create a model that determines if coffee is good or bad. Sure, that's even more arbitrary, but it's more useful. Then we could take samples from a bunch of different coffee shops, and finally scientifically prove that Starbucks' coffee blows.



Lindinger, C., Labbe, D., Pollien, P., Rytz, A., Juillerat, M., Yeretzian, C., Blank, I. (2008). When Machine Tastes Coffee: Instrumental Approach To Predict the Sensory Profile of Espresso Coffee. Analytical Chemistry, 80(5), 1574-1581. DOI: 10.1021/ac702196z

Photo: Louisiana State Museum
Photo: Flickr

Godspeed.


This posted tagged as: food, journals, science, coffee

Recent posts tagged as journals:

• Quadruple-clicked Catenanes
• Slushes, Nanowords and Footnotes
• Cool abstract picture of the day and a Lack of binding

#1: by joel on 05/21/08 @ 5:36PM EST

Rad paper. I'd be interested in hearing how they pulled their espresso shots (packing density temperature can effect the flavour as much as extraction time, though I'm guessig they pulled them equally for each sample(?)), and the kind of grinder they had.

...err, can you tell I spend too much time at the coffee shop?

Reply to this comment.

#2: by will on 05/21/08 @ 5:59PM EST

Hmm.. yeah I couldn't find that in the paper. As for the grinder, they don't say, they just say that it's "A standard espresso machine (TURMIX C250, Zurich, Switzerland) was used for the preparation of espresso coffee."

Unless it has a built-in grinder, it doesn't say.

Haha, nothing wrong with lots of time at the coffee shop; we all know that coffee is the fuel of science.

Reply to this comment.

#3: by Yeretzian on 11/08/08 @ 3:35PM EST

We used Nespresso capsules.
So no need for grinding. The coffee you see in the paper are the different varietes of Nespresso.

Name:


Website:


Enter a valid email:

(Gravatars and retaggr profiles are supported.)

Receive notification of followup comments?
Yes No

Comment:
(First time commenters: Don't use <a href> tags or you'll be flagged as spam.)


Reply to main post. (To reply to an individual comment, select it above.)