jfouyang
commented
3 years ago Hi, that's a very nice customized version of ShinyCell app! :)
If you look into server.R, this chunk of code controls the colour palette for gene co-expression plot:
# Generate coex color palette
cInp = strsplit(inpcol, "; ")[[1]]
if(cInp[1] == "Red (Gene1)"){
c10 = c(255,0,0)
} else if(cInp[1] == "Orange (Gene1)"){
c10 = c(255,140,0)
} else {
c10 = c(0,255,0)
}
if(cInp[2] == "Green (Gene2)"){
c01 = c(0,255,0)
} else {
c01 = c(0,0,255)
}
c00 = c(217,217,217) ; c11 = c10 + c01
nGrid = 16; nPad = 2; nTot = nGrid + nPad * 2
gg = data.table(v1 = rep(0:nTot,nTot+1), v2 = sort(rep(0:nTot,nTot+1)))
gg$vv1 = gg$v1 - nPad ; gg[vv1 < 0]$vv1 = 0; gg[vv1 > nGrid]$vv1 = nGrid
gg$vv2 = gg$v2 - nPad ; gg[vv2 < 0]$vv2 = 0; gg[vv2 > nGrid]$vv2 = nGrid
gg$cR = bilinear(gg$vv1, gg$vv2, nGrid, c00[1], c10[1], c01[1], c11[1])
gg$cG = bilinear(gg$vv1, gg$vv2, nGrid, c00[2], c10[2], c01[2], c11[2])
gg$cB = bilinear(gg$vv1, gg$vv2, nGrid, c00[3], c10[3], c01[3], c11[3])
gg$cMix = rgb(gg$cR, gg$cG, gg$cB, maxColorValue = 255)
gg = gg[, c("v1", "v2", "cMix")] To summarise briefly,
First, these four variables define the extreme colours
c00: colour for zero expression of both gene A and B, which is set to RGB of c(217,217,217)
c10: colour for max expression of A, zero expression of B
c01: colour for zero expression of A, max expression of B
c11: colour for max expression of both gene A and B, which is set to c10 + c01
Second, we then define the number of bins to bin the colours, which is set to 16 bins (nGrid = 16) and we also added some "padding" on both the lower and higher end of the expression nPad = 2. So there will be a total of 20 bins (2+16+2) where the first 2 bins, as well as the third bin, are of the same colour. This basically adds the grey bits to the bottom left side of the colour palette.
Third and finally, the whole colour palette is generated using a bilinear interpolation function.
I think what you want is to make grey part of the bottom left side of the colour palette extend further. This would require you to modify nPad in some way. Maybe you can try the following code and change the nPadLow?:
nGrid = 16; nPadLow = 2; nPadHigh = 2; nTot = nGrid + nPadLow + nPadHigh
gg = data.table(v1 = rep(0:nTot,nTot+1), v2 = sort(rep(0:nTot,nTot+1)))
gg$vv1 = gg$v1 - nPadLow ; gg[vv1 < 0]$vv1 = 0; gg[vv1 > nGrid]$vv1 = nGrid
gg$vv2 = gg$v2 - nPadLow ; gg[vv2 < 0]$vv2 = 0; gg[vv2 > nGrid]$vv2 = nGrid And one last important thing! This chunk of code appears twice, once for the main plot, once for the legend so the code needs to be replaced twice.
Hope that my explanation is not too confusing!
jvelghe
Hi John, I'm currently tweaking some items on the application, and was wondering if it is possible to adjust the colour threshold for the gene co-expression plots? We changed the threshold for some Seurat plots and were hoping to do something similar for this interface as well to make it more biologically representative.
Along with two Seurat co-expression examples, here is the work-in-progress link to the app for your reference: https://lynnlab.shinyapps.io/Human_Islet_Atlas/
INS_GCG_Endo_FeatBlend.pdf INS_SST_Endo_FeatBlend.pdf
If you could point me in the direction of where to modify the server code I would be very appreciative :)
Thank you!