feat: restyle to Sina plots and generate manuscript panel figure.

biosustain / memote-meta-study

Test metabolic models in the wild and rank them according to popularity and how easily they can be improved.

Apache License 2.0

3 stars 2 forks source link

feat: restyle to Sina plots and generate manuscript panel figure. #1

Closed ChristianLieven closed 5 years ago

ChristianLieven commented 5 years ago

Hey @Midnighter and @phantomas1234 ,

in this PR I made the following changes:

Added a SinaPlot layer (https://doi-org.proxy.findit.dtu.dk/10.1080/10618600.2017.1366914) but configured it not to scale the width of all collections by the collection with the highest density!
Added y-axis labels to all the individual figures that are more descriptive than simply score.
I only inverse the metric (aka. 1 - metric) for scored tests. For unscored tests the metric will no longer be switched around. This may be in conflict with the added y-axis labels, which will have to be double-checked.
Added plot_figure.R which combines a selection of 3 individual figures and one clustering figure into a "publication ready [?!]" panel.
Arranged the x-axis labels by Automatic: Agora, CarveMe, KBase, Path2Models and Manual: BiGG, Ebrahim and U. Minho, with these labels instead of the column names.

ChristianLieven commented 5 years ago

Pushed all the figures on request from @phantomas1234
Removed the x-axis labels on the two center panels in the final figure.

Midnighter commented 5 years ago

I only inverse the metric (aka. 1 - metric) for scored tests. For unscored tests the metric will no longer be switched around. This may be in conflict with the added y-axis labels, which will have to be double-checked.

Why did you decide to do this? Won't it confuse readers that the scale is suddenly inversed?

Midnighter commented 5 years ago

Just looked at the panel figure. Must admit I'm not happy with the y-axis label wrapping over 4 lines. At least for the panel I'll play around a bit more.

ChristianLieven commented 5 years ago

Just looked at the panel figure. Must admit I'm not happy with the y-axis label wrapping over 4 lines. At least for the panel I'll play around a bit more.

You can change that easily by manipulating the width parameter in this line: ylabels = str_wrap(y_axis_labels[as.character(test)], width = 25),

I only inverse the metric (aka. 1 - metric) for scored tests. For unscored tests the metric will no longer be switched around. This may be in conflict with the added y-axis labels, which will have to be double-checked.

Why did you decide to do this? Won't it confuse readers that the scale is suddenly inversed?

I thought having the same or similar y-axis labels as the Test names in our report would make sense.