Skip to content

Latest commit

 

History

History
92 lines (58 loc) · 3.08 KB

README.md

File metadata and controls

92 lines (58 loc) · 3.08 KB
Raw

Recovering origin and ratings of Bordeaux wines with raw gas chromatography

All figures can be reproduced using the following three scripts:

  • estate_vintage_clf.py: Decoding estate and vintage from wine chromatograms
  • parker_clf.py: Decoding Parker's rating from wine chromatograms
  • embed.py: Embedding wine chromatograms in 2d (3d) space using umap and tSNE

Each script contains functions to process raw data that is found in Source_Data/data resulting in files for plotting, stored in Source_Data/res, precomputed for your convenience. Running all analyses from the raw data takes about 3 h on a laptop in total. All figures are precomputed in Source_Data/figs.

The Source_Data/data contains the following:

Gas chromatography features per wine, three types and concatenation

  • oak.npy, 80 wines x 30275 GC features, each wine e.g. with name A_2012, i.e. estate A, vintage 2012
  • esters.npy, 80x7881
  • offFla.npy, 80x10480
  • concat.npy, 80x48636

Manually selected compounds of gas chromatography per wine, three types and concatenation

  • m_oak.npy, 80x13
  • m_esters.npy, 80x16
  • m_offFla.npy, 80x3
  • m_concat.npy, 80x32

Additional wine info

  • 32_compound_names.npy, chemical names for 32 compounds
  • parker_ratings.npy, 80x1, Parker ratings (int between 0 and 100)
  • varietals.npy, 80x4, percentage of 4 grape types per wine
  • bordeaux_map.png, map of Bordeaux area

Prerequisites

In your python environment (for example installed via anaconda, latest version) install required libraries:

pip install --requirement requirements.txt

List of figures with function name and script name to reproduce:

  • Figure 1 plot_tile_main() in embed.py
  • Figure 2, 5(c,d) plot_violin() in estate_vintage_clf.py
  • Figure 3, S9 plot_chunks2() and plot_all_survival() in estate_vintage_clf.py
  • Figure 4, Table S1, Figure 5e,f plot_violin_p() in parker_clf.py
  • Figure 5a,b plot_tile_chem32() in embed.py
  • Figure S2 plot_tile_supp() in embed.py
  • Figure S3 plot_best_concat() in estate_vintage_clf.py
  • Figure S4 plot_vintage_decoding_per_wine() in estate_vintage_clf.py
  • Figure S5 PCA_features() in estate_vintage_clf.py
  • Figure S6 plot_weights() in estate_vintage_clf.py
  • Figure S7, S8, S11 plot_all_survival() in estate_vintage_clf.py
  • Figure S12 plot_weights_m_abs() in estate_vintage_clf.py
  • Figure S10 plot_chunk2() in parker_clf.py
  • Figure S13 plot_chunks_m_dists() in estate_vintage_clf.py
  • Table S2 varietals_table() in embed.py

Example application:

To clone this repository into your home directory, type in a terminal:

git clone https://github.com/mschart/wine_decoding.git

To reproduce figure 1, navigate into directory wine_decoding, start an ipython session and type:

# read in script (pip install missing libraries if needed) for figure 1
run '/home/mic/wine_decoding/embed.py'

# activate interactive plotting
plt.ion()

# plot figure 1 by running listed function (loading in raw data directly)
plot_tile_main()

Some figures need manual assembly (figure 5, 3, S2).