R circlize chord diagram with log scale

问题

Is it possible to make a chord diagram in the circlize package that displays log10 values? So far I have been able to produce a plot with correct size links, but the corresponding axis will not match up. the axis displays the sum of all links/logged values per sector, which is not correct as summing logged values does not correspond to summed raw values. Is there any way to fix this axis problem?

below is an example of what I have tried so far

library(circlize)

export_country <- c("DEU","USA","IDN","USA","IDN","USA","IDN","CAN","DEU","DEU","IDN","NZL","DEU","USA","USA","USA","IDN","SGP","IDN")
import_country <- c("JPN","JPN","USA","JPN","TWN","CAN","CHN","USA","CHN","CHN","DEU","JPN","USA","DNK","JPN","CHN","JPN","CHN","CHN")
flow <- c(2000,65780,78010,851,35353,845,738,120788,245900,90002,4426,6870,152681,78114,32591,19274,10915,23100,6275)

df <- data.frame(export_country, import_country, flow,stringsAsFactors = FALSE)


country = unique(c(df[[1]], df[[2]]))
color <- c("#E41A1C","#800000","#ff8c00","#ffd700","#008000","#00bfff","#377EB8",
               "#ff69b4","#800080","#4b0082")

df1 <- data.frame(country, color,stringsAsFactors = FALSE)

circos.clear()
circos.par(start.degree = 90, gap.degree = 5, track.margin = c(-0.1, 0.1), points.overflow.warning = FALSE)
par(mar = rep(0, 4))

chordDiagram(x = df[1:2],log10(df[3]), grid.col = color, transparency = 0.25,
         order = country, directional = 1,
         direction.type = c("arrows", "diffHeight"), diffHeight  = -0.04,
         annotationTrack = c("grid","axis"), annotationTrackHeight = c(0.05, 0.1),
         link.arr.type = "big.arrow", link.sort = TRUE, link.largest.ontop = TRUE)



circos.trackPlotRegion(
  track.index = 1, 
  bg.border = NA, 
  panel.fun = function(x, y) {
xlim = get.cell.meta.data("xlim")
sector.index = get.cell.meta.data("sector.index")
country = df1$country[df1$country == sector.index]

circos.text(x = mean(xlim), y = 4.4, 
            labels = country, facing = "bending", cex = 1, niceFacing = TRUE, adj = c(0.5, 0))

  }
 )

which gives the following plot

回答1:

I think it is impossible to solve this problem. Since a sector is composed by several links, if the size of the sector is log-transformed, what does the width of each link mean? I think we should better forget the scale of each sector and do not show the axes. On the other hand, we can directly show the un-log transformed value beneath or above each link.

In following code, actually, chordDiagram() returns a data frame which contains positions of each link, then we can use this information to add un-logged values just to the right place.

Also please note the first argument in chordDiagram() in your code was wrong. I corrected it.

library(circlize)

export_country <- c("DEU","USA","IDN","USA","IDN","USA","IDN","CAN","DEU","DEU","IDN","NZL","DEU","USA","USA","USA","IDN","SGP","IDN")
import_country <- c("JPN","JPN","USA","JPN","TWN","CAN","CHN","USA","CHN","CHN","DEU","JPN","USA","DNK","JPN","CHN","JPN","CHN","CHN")
flow <- c(2000,65780,78010,851,35353,845,738,120788,245900,90002,4426,6870,152681,78114,32591,19274,10915,23100,6275)

df <- data.frame(export_country, import_country, flow,stringsAsFactors = FALSE)
df[[3]] = log10(df[[3]])

country = unique(c(df[[1]], df[[2]]))
color <- c("#E41A1C","#800000","#ff8c00","#ffd700","#008000","#00bfff","#377EB8",
               "#ff69b4","#800080","#4b0082")

df1 <- data.frame(country, color,stringsAsFactors = FALSE)

circos.clear()
circos.par(start.degree = 90, gap.degree = 5, track.margin = c(-0.1, 0.1), points.overflow.warning = FALSE)
par(mar = rep(0, 4))

res = chordDiagram(x = df, grid.col = color, transparency = 0.25,
         order = country, directional = 1,
         direction.type = c("arrows", "diffHeight"), diffHeight  = -0.04,
         annotationTrack = c("grid"), annotationTrackHeight = c(0.05, 0.1),
         link.arr.type = "big.arrow", link.sort = TRUE, link.largest.ontop = TRUE)



circos.trackPlotRegion(
  track.index = 1, 
  bg.border = NA, 
  panel.fun = function(x, y) {
    xlim = get.cell.meta.data("xlim")
    sector.index = get.cell.meta.data("sector.index")
    country = df1$country[df1$country == sector.index]

    circos.text(x = mean(xlim), y = 1.5, 
                labels = country, facing = "bending", adj = c(0.5, 0), cex = 1, niceFacing = TRUE)

      }
 )

for(i in seq_len(nrow(res))) {
  circos.text(x = res$x1[i] - res$value[i]/2, y = 0.5, round(10^(res$value[i])), facing = "inside",
      niceFacing = TRUE, adj = c(0.5, 0.5), cex = 0.5, col = "white", sector.index = res$rn[i], track.index = 1)
  circos.text(x = res$x2[i] - res$value[i]/2, y = 0.5, round(10^(res$value[i])), facing = "inside",
      niceFacing = TRUE, adj = c(0.5, 0.5), cex = 0.5, col = "white", sector.index = res$cn[i], track.index = 1)
}

来源：https://stackoverflow.com/questions/39504096/r-circlize-chord-diagram-with-log-scale

标签

logging

chord-diagram

circlize