Tuesday, October 18, 2016

The Genome Cellar is no such thing

In an earlier blog post, I noted that The Music Genome Project is no such thing. The use of the word "genome" in this context is an analogy, in which the musical characteristics are seen as producing a sort of genetic fingerprint. However, this is a false analogy, because the data used for the Music Genome Project are actually phenotypic, not genotypic. Indeed, music has no analog of a genotype.

In a similar vein, the data used for The Genome Cellar are phenotypic, not genotypic, and so this is also a false analogy.

The Genome Cellar is the database used by the Next Glass app. This app was released in November 2014, and a concurrent press release explained the concept:
Next Glass is the breakthrough app that uses science and machine learning software to provide accurate, personalized recommendations to consumers. Next Glass has analyzed tens of thousands of bottles of wine and beer with a mass spectrometer and stores the "DNA" of each product in its Genome Cellar™, which combines with users' Taste Profiles™ to provide product-specific recommendations.
So, the beer / wine data in the Genome Cellar are peaks in a spectrophotometer output. This is made clear in another press release:
Next Glass has developed the world’s first Genome Cellar, an extensive database that contains the chemical makeup – or "DNA" – of tens of thousands of wines and beers. By looking at each bottle on a molecular level, Next Glass defines a unique taste profile for every bottle by analyzing thousands of chemical elements.
This procedure will, indeed, provide a unique fingerprint for each alcoholic product, but it will be a phenotypic one not a genotypic one. Genetics is often chemistry but not all chemistry is genetics.

The idea of the Next Glass app is the same as that for the Music Genome Project — to use the fingerprint of currently liked products (music or wines / beers) to make recommendations for other products that might appeal to the customer. This approach can be expected to work for alcoholic beverages, because the subjective preferences will be based to some extent on the sensory components of the chemical makeup. If you document enough of the chemistry then you are bound to include a large proportion of the sensory part.

Anyway, you can see a short video about the laboratory here.

Finally, you might like to compare this approach with that of WineFriend, which tries to assess your taste in wine with multiple-choice questions, instead of complex chemistry. WineFriend:
uses a simple eight question taste survey that gives insights into a customer's thresholds for sweet, sour, bitterness and intensity of flavour. It then creates a profile which enables it to select wines that are tailored to the individual customer's tastes.
No mention of genomes here.