In the ever increasing trend to treat beer more like wine, there are now dozens of suggestions online for beer / food pairings. There are some serious and useful (don’t get me wrong, I like the food pairing stuff) graphics from places like craftbeer.com to help with this. As far as I can tell, the way to determine if a liquid and a food go together is to collect anecdotes, the lowest form of evidence from a scientific point-of-view.

“Are people doing it currently? Does it make the two things taste better? Then it is a food pairing.” Not a very thoughtful way to determining something.

I’ve been increasingly curious with why things taste good together. I’ve had discussions that do amount to “well I know better” about beer and food pairing, specifically about whether or not beer pairs well with spicy foods. I decided to look into it the way an academic would look into it. I don’t want to take anything away from people like Garret Oliver who wrote one of my favorite books on the subject or Greg Engert, owner of one of my favorite bars. While their opinions are useful and definitely valid, they ultimately boil down to a sort of “argument from authority”…I’m right because I know I’m right sort of thing. Also, advice is separate from measurement and science. I want to know the science behind taste perception, not just looking at peoples advice. There are a dozen blogs on my feed filled with beer / food recipes that are full of good advice. To my pleasant surprise, there is some actual science behind food pairing and some people doing legitimate work to discover not only WHAT foods go together but WHY foods go together.

It turns out that food pairing is studied scientifically and one of the leading hypotheses about how to pair foods is to look at what aroma/flavor chemicals are shared between the two substances. On the surface it makes sense in a “why didn’t I think of that” sort of way. Like goes with like, but more at the chemical level. Pairing two foods that have the same chemical make-up, or the same array of flavor-active compounds, might enhance the flavors, it makes sense at least. What does this observation means for “beer” as a food product.

Obviously, beer is a complicated food product. Beer is a beverage that combines the flavors of primary beer ingredients (malt, water, hops, and yeast) as well as what other spice or adjunct you added to the beer during or after fermentation (coriander, citrus, juices, fruits, honeys, etc). On top of this complexity there are more than 100 different KINDS of beer, and the flavor components of those different styles of beers will obviously affect the flavor and therefore the pairing.

So I was poking around on the internet the other day and came across this article from 2011. I thought it was interesting enough to summarize and elaborate on it here. The authors of the article built an impressive database of flavor/aroma active chemicals and assigned them to different foods, then drew association maps between different food products and showed that foods that shared more chemicals were classically considered to be paired together. Is this definitive? No, absolutely not, but it is a fascinating way to look at food. The basic premise of the article is that chemicals form groups based on what flavor sensations they evoke within your sensory system, chemicals that are similar or the same will generate similar perceptions and so they will pair together. The graphic belwo shows what chemicals are shared in a dish of mussels with a tomato sauce. There are dozens of shared chemical components with taste consequences.

flavorchemicals1
Figure from Ahn et al 2011 in Nature

“Cool, you just told me why spaghetti sauce tastes good….neat…” OK OK, this is a BEER blog. Let’s talk about beer.

I’ve touched on some of the chemicals that affect flavors that are produced during a normal fermentation with Saccharomyces species and with Lactobacillus species. But those posts focused primarily on what flavors the yeast or bacteria itself produces, not what flavors the hops or the malts would produce independent of the fermentation. I was curious what flavor compounds were used in this analysis since “beer” was included in the food products. To a beer snob like myself, calling something “beer” raises a lot of questions. It is sort of like calling a food “cheese”…well it’s a place to start but a pilsner and a stout are both beer and both chemically distinct not to even mention mixed fermentation beers that will contain lactic acid as a major product. “Beer”….I left a table at the end of this document with the 229 compounds found in beer (a pale ale) used to make this network of flavor interactions (there were a total of 1106 compounds in the database used) .

net
Figure from Ahn et al 2011 in Nature

That is a busy graphic and hard to read but allow me to summarize it for you. Beer shared at least one compound with 948 different food products (out of 1532 food products in the database).  Their analysis of shared compounds between beer and their panel revealed that beer had the most in common with the following Blue Cheese, Cheddar Cheese, Swiss cheese, Gruyere cheese, Beef, Apple, Pork, Liver, Katsuobushi, Milk, Roast, Beef, Coffee, Corn, Cocoa, Galanga, Soy Sauce, Watercress, and Rutabaga. Some of these are not surprising. Really quick, Katsuobushi is dried fermented tuna. I had never heard of it, perhaps I should get out more. Without lumping or editing their database, I pulled out the top 75 hits for “beer” and flavor pairings, table below.

Food Product # of Shared Compounds
white wine 118
cabernet sauvignon wine 107
red wine 106
roasted beef 106
jamaican rum 104
rum 103
boiled beef 103
sherry 103
fried beef 103
braised beef 102
blue cheese 102
grilled beef 102
raw beef 102
coffee 101
gruyere cheese 101
parmesan cheese 100
cognac 99
cheddar cheese 97
black tea 97
roasted cocoa 96
roasted green tea 95
green tea 95
swiss cheese 95
cocoa 94
camembert cheese 92
cooked apple 92
sparkling wine 91
provolone cheese 91
botrytized wine 90
mozzarella cheese 90
port wine 90
fermented tea 90
feta cheese 89
rose wine 89
rooibus tea 89
russian cheese 89
domiati cheese 88
dried green tea 88
romano cheese 88
dried black tea 88
limburger cheese 88
brewed tea 88
cottage cheese 88
jasmine tea 88
tea 88
pouching tea 88
ceylon tea 88
seychelles tea 88
kola tea 88
fermented russian black tea 88
comte cheese 88
strawberry 87
sheep cheese 87
cheese 87
soybean 87
cream cheese 87
goat cheese 87
roquefort cheese 87
tilsit cheese 87
munster cheese 87
emmental cheese 87
wine 86
champagne wine 86
peanut butter 82
roasted peanut 82
cured pork 81
roasted chicken 79
milk 78
butter 77
grilled pork 73
tomato 73
uncured smoked pork 72
roasted pork 72
beer 70

There are definitely themes in this list. Such as….wine, beef, tea, and cheese. Also known as the best foods. So not surprising, there are a lot of fermented foods on that list. A lot. What are the flavor chemicals that are driving those interactions exactly? Well, there are several chemicals that appear in dairy, meat, and beer products. The chemicals driving this interaction include o-cresol, 2-butanone, propyl acetate, ethyl lactate, decanoic acid, isoamyl laurate, pyrazine, phenethylamine, thiazole, indole, pyrrolidine, 3-hydroxy-2-pentanone, 2-hepten-4-one, ethyl isovalerate, propyl acetate, n-valeraldehyde, methyl acetate, and guaiacol. Looking at a list of chemicals like that makes you wonder how these lists were made and if they are weighted at all. I traded emails with the authors of the paper and as far as I can tell, the different chemicals were not weighted (or were weighted the same if you prefer) and the lists were populated from the literature and a couple of books that contained lists of compounds found in food products.

If you like beer, eat chicken, beef, pork, and dairy products is the bottom line. Also tea…I’m fine with all of that.

Final Thought

This sort of analysis is a deep rabbit hole. There are a lot of limitations to this sort of thinking and analysis, most notably is you are limited by what compounds you include in the initial analysis and I was initially surprised that ethanol was not on their list (especially since even without it all the alcoholic beverages still clustered together). Ethanol definitely has a flavor and would influence this sort of analysis as well as people’s taste preferences but I understand how that is such a huge contributor that you might not want it in the analysis. It would also affect the solubility of other chemicals that would have flavors…enhancing those. Second, what foods you select are going to bias your analysis. In this respect I think the authors did an amazing job selecting a variety of foods from several cultures that hit several “areas” of food. At times I found the list a little redundant but I should never judge their database building. It would be great to have a database that was 10 or 100x larger though, then you could really do some detailed analysis. Lastly, there wasn’t a serious effort to take into consideration the overall contribution each flavor compound had on the individual food or drink item. The threshold for different taste chemicals would be different and the overall taste detection would be dependent on the concentration of each compound. This would be really hard to do though and I recognize that. Also, it might be misleading as lots of things that do pair do NOT share a compound that is dominant in one of the ingredients. It would be pretty hard to parse that sort of data in a meaningful way.

Like I said above, it would be interesting to do this with a few different kinds of beers. I didn’t include it in the analysis for a couple of reasons but the authors included a second beer they labeled as “bantu beer”…I did not include it in the analysis above because it was so close to the generic beer, they shared the greatest number of compounds of ANY TWO entries in the entire database at 227, grilled beef and roast beef were second at 207 shared compounds. Bantu beer and beer were nearly interchangeable in this analysis, often appearing side-by-side in the cells of the spreadsheet. In fact, the primary difference between the two beers on their list came down to the presence of three simple compounds in the bantu beer not listed in the “beer”: ethylene oxide, methyl myristate, and 1-phenyl-1-propanol. I’m not sure why those compounds were listed with bantu beer and not with the generic beer but perhaps they are present due to the sorghum used in the fermentation. Who knows? Not me. It would be fun to do a stout versus a pale ale or a lager versus an ale though. Perhaps if I have a bottomless pit of time I will do that analysis one day.

Ultimately, one could spend hours pouring over their data and making interesting connections. My primary interest in this centers around the fact that this sort of analysis has the potential to work. Provided you could side step the obvious problems I pointed out in the beginning with chemical

 

 

“Across the troubled maelstrom of time, people always need beer.” -Ellen Kushner

 

 Supplemental Information. List of compounds found in beer used for the matching.

jasmone
methyl 2-methylthiobutyrate
methylsulfinylmethane
benzothiazole
2-acetylfuran
nonyl acetate
4-hydroxy-3-methoxybenzoic acid
2-methylpyrazine
p-ethylphenol
isoamyl formate
3-ethylpyridine
isoamyl acetate
4-methyl-2,3-pentanedione
g-dodecalactone
isoamyl octanoate
2,3-dimethylpyrazine
ethyl butyrate
g-hexalactone
ethyl-3-methylthiopropionate
pyruvic acid
methyl-2-pyrrolyl ketone
3-methylbutanethiol
2-methylbutyl acetate
2-nonanone
4-decenoic acid
isoamyl propionate
isopentylamine
isobutyl formate
oleic acid
cis-3-hexenal
eugenol
1-decanol
ethyl isobutyrate
2-ethyl-3,5(6)-dimethylpyrazine
undecanal
valeric acid
amyl butyrate
resorcinol
1-octen-3-ol
2-acetyl-5-methylfuran
2-ethyl-5-methylpyrazine
2-hexen-1-ol
palmitic acid
2-nonanol
g-decalactone
hexanoic acid
pyridine
isoamyl benzoate
ethyl 3-hexenoate
myristaldehyde
2,4-dihydroxybenzoic acid
isoamyl hexanoate
3-hexanone
lauric acid
ethyl lactate
vanillin
3-(methylthio) propionaldehyde
s-methyl 4-methylpentanethioate
ethyl thioacetate
1-octanol
nonanoic acid
ethyl octanoate
2-octanol
n-furfuryl pyrrole
ethyl acetate
isobutyl benzoate
isobutyl acetate
furfuryl alcohol
g-nonalactone
4-(2,6,6-trimethyl-cyclohexa-1,3-dienyl)but-2-en-4-one
cinnamic acid
isobutyric acid
decanoic acid
hexyl octanoate
pyruvaldehyde
2-oxobutyric acid
2-methylbutyric acid
d-camphor
6-methyl-5-hepten-2-one
g-valerolactone
heptyl alcohol
isoamyl laurate
methyl nicotinate
2-trans, 4-trans-decadienal
nona-2-trans,-6-cis-dienal
4-hydroxy-3,5-dimethoxybenzaldehyde
vanillin
4-methyl-2,6-dimethoxyphenol
pyrrole
propyl mercaptan
9,12-octadecadienoic acid (48%) plus 9,12,15-octadeca- trienoinc acid (52%) (methyl esters)
nerolidol
2-methylpentanal
prenylthiol
(+/?) heptan-3-yl butyrate
heptyl acetate
g-butyrolactone
hydrogen sulfide
4-mercapto-4-methyl-2-pentanone
ethyl propionate
1-butanethiol
citronellyl acetate
ethyl nonanoate
octanoic acid
diethyl sulfide
butylamine
amyl hexanoate
phenethyl butyrate
hexyl isobutyrate
isovaleric acid
2-ethylpyrazine
n-nonanal
4-methyl-2-pentanone
dimethyl trisulfide
isopropyl acetate
(+/?)-2-(5-methyl-5-vinyltetrahydrofuran-2-yl)propi-onaldehyde
hexyl butyrate
2,6-dimethylpyridine
2-octanone
methyl sulfide
propyl alcohol
phenethyl acetate
hexyl acetate
lauryl alcohol
4-hydroxybenzoic acid
2-hexenal
2-methyl-1-propanethiol
4-ethylguaiacol
3-(methylthio)propyl acetate
n-octanal
2-ethyl-3-methylpyrazine
decanal
myrcene
4-hydroxybenzaldehyde
trimethylamine
4-hydroxy-2,5-dimethyl-3(2h)-furanone
2-methyltetrahydrothiophen-3-one
alpha-terpineol
3-hydroxy-2-oxopropionic acid
phenethyl propionate
geranyl isobutyrate
propionic acid
3-acetylpyridine
heptanoic acid
3-methyl-2-oxobutanoic acid
3-(methylthio)propanol
lauric aldehyde
phenylacetaldehyde
phenylacetic acid
methyl mercaptan
ethyl heptanoate
3-phenylpropionic acid
2,3,5,6-tetramethylpyrazine
2-methylbutyraldehyde
benzyl isobutyrate
fenchyl alcohol
2-hydroxybenzoic acid
3-methylbutyraldehyde
octyl butyrate
2-trans, 6-trans-nonadienal
4-ethyl-2,6-dimethoxyphenol
geranyl acetate
1-penten-3-ol
a-phellandrene
ethyl formate
isobutyl alcohol
isobutyl hexanoate
myristic acid
5- and 6-decenoic acid
salicylaldehyde
2-methyl-3-butenal
2-methoxy-4-vinylphenol
pyrazine
isoamyl nonanoate
phenethylamine
heptyl formate
maltol
5h-5-methyl-6,7-dihydrocyclopenta(b)pyrazine
l-malic acid
2-methoxy-4-methylphenol
3-phenylpropionaldehyde
2,3,5-trimethylpyrazine
2,3-pentanedione
thiazole
3-methylbutyl-2-methylpropanoate
2-heptanone
2-acetylpyridine
indole
n-butyric acid
pyrrolidine
2-pentanone
p-vinylphenol
3-hydroxy-2-pentanone
furfuryl acetate
glycerol
3-methyl-2-buten-1-ol
phenethyl hexanoate
2-acetylthiazole
4-propenyl-2,6-dimethoxyphenol
methyl hexanoate
2-pentanol
2-hepten-4-one
methyl disulfide
phenethyl isovalerate
ethyl isovalerate
5-methylfurfural
diacetyl
citronellol
propyl acetate
4,5-dimethyl-3-hydroxy-2,5-dihydrofuran-2-one
propionaldehyde
carvacrol
n-valeraldehyde
nerol
2-methyl-2-pentenal
methyl acetate
3-hexenoic acid
phenethyl alcohol
isopropyl alcohol
phenethyl octanoate
phenol
s-methyl thioacetate
phenethyl isobutyrate
nonyl alcohol
s-methyl hexanethioate
2-heptanol
guaiacol
Advertisements

7 thoughts on “Beer Pedantry #3: Beer Flavor Pairings (science based rant)

  1. Did they assess whether or not these compounds were above threshold or not? The info is kind of useless for below threshold sensory compounds. It’s noise. More interesting to compare quantitative data that is above a qualitative threshold, to the compare lager to stout to blur cheese to roast beef.

    1. Supposedly, the lists were compounds that were above the threshold but I touched on that concern in my wrap up. Obviously taking into consideration the different concentrations would be better but far more difficult to do.

  2. Interesting article! Olfactory experience would also seem to be important and it would add another level of complexity. “Things are always more complicated than they are.” 🙂

  3. I really like all of the questions that the referenced study raises. It would seem that there is a database out there of various flavor and aroma compounds in foods, and at what quantity. If a flavor threshold was assigned to each compound, a more meaningful relationship could be found. This doesn’t sound ridiculously difficult, but it assumes said data exists. Of course, there are a lot of other factors at play, such as which grouping of flavors work best together, factoring in customary ingredients (salt on eggs, rosemary on a roast) and textures (carbonation, viscocity, creaminess, etc.) that are often part of the experience with each food item, as well as how compounds are perceived at various levels, but it would be a decent starting ground.

  4. It seems like this kind of approach could also help to come up with new ideas for beer adjuncts. I’m not sure how you’d do a cheese or beef beer, but coffee is already popular and I’ve made a galangal ale before that was really good. I’ve seen very few beers that use tea as an adjunct, so that seems like an area ripe for experimentation.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s