Feature request: map multiple numerical and categorical data to plots

[ljournot]

I am 1) a newbie to Julia (and to programming) 2) a biologist 3) French. It's tough but c'est la vie !
In biology we often want to map several kind of numerical (e.g. fold enrichment, gene counts, p-value) and categorical (e.g. gene ontology category) data in a single plot. Some colleagues of mine produced the attached plot using R and ggplot2. I failed to produce a similar plot using Gadfly. I could map the p-value to the point color (Scale.ContinuousColorScale(Scale.lab_gradient(colorant"white",colorant"orange",colorant"red"))), but could not change their shape and size to map GO category and gene count, respectively. I suspect a solution might be something like change the Theme for each point, but it sounds quite complicated and I still don't find out how to change the shape depending on categorical values. Any idea ?

Thanks a lot for your help.

laurent

[tbreloff]

I might be able to help, but it would be much easier if you supplied your data. Can you copy it into into a github gist?

[tbreloff]

So I took a stab at this using Plots.jl (honestly, I don't plan on attempting this in pure Gadfly, I think it will be too annoying)

Feel free to open issues at https://github.com/tbreloff/Plots.jl/issues with questions/requests.

[ljournot]

Hi Tom. Thanks for taking care of my request. I will look at Plots over the next few days. Just in case you want to go on, please find attached our data. laurent

Data.txt

[ljournot]

it looks like the Gadfly backend used by Plots has an intrinsic problem with variable marker size.

Here is my script:
using DataFrames
cd("C:/Users/Laurent/Documents/LAURENT/MANIPS/Julia scripts/FairyLights")
data=readtable("Input//Data.csv");
using Plots
gadfly()
dataframes()
plot(data,:Fold_Enrichment,:Term,group=:Category,size=(500,500),
m=([:r :c :V :d],data[:Gene_Count]/10),mc=[RGB{U8}(1,0,0) RGB{U8}(0,1,0) RGB{U8}(0,0,1) RGB{U8}(1,0,1)],
)

and here is the message I got 4 times (one for each Category):
WARNING: Gadfly doesn't support variable marker size... using the average: 1.902325581395

The plot is working with a fixed marker size.

[Mattriks]

This is an update. I've reproduced the above plot using Gadfly master: except for the size and shape scales - which hopefully will be a future development! Note I'm not sure if size was linked to area or radius in the plot above.

using DataFrames, Gadfly
# There is a link to Data.txt in the issue above
D = readtable("Data.txt", header=true)
D[D[:Weight_Fischer].>0.01, :Weight_Fischer] = 0.01

theme1 = Theme(point_size_min=2pt, point_size_max=9pt, 
    point_shapes=[square, circle, utriangle, diamond])

shapesc = Scale.shape_discrete(levels=["BP","CC","KEGG_PATHWAY","MF"])
colsc = Scale.color_continuous(colormap = Scale.lab_gradient(colorant"red",
                                colorant"orange", colorant"yellow"))

p = plot(D[end:-1:1,:], x=:Fold_Enrichment, y=:Term,  size=:Gene_Count, shape=:Category,
            color=:Weight_Fischer, Geom.point, theme1, shapesc, colsc,
            Coord.cartesian(xmin=0, xmax=22),
            Guide.colorkey("P-value") 
)