8 — Recipes

18. Circular Stacked Bar Plots

This tutorial shows how to use Circos to create a general information graphic, not related to genome visualization. Or at least, not obviously related.

The image is a series of concentric stacked bar plots. Each ring corresponds to several data points, which have been normalized so that their sum is constant (e.g. 1000). We'll hide the scale and ideograms, so only the data appear.

Another example of the use is Circos for illustration is described in the Nature Cover tutorial, which shows you how to recreate the Encode Nature cover.

creating the axis

The axis is a single "ideogram" of length 1000. We'll normalize all the data sets to add up to 1000. The axis name is gh.

# scale.txt
chr - gh gh 0 1000 black

The karyotype is set to this file

# circos.conf
...
karyotype                   = scale.txt
chromosomes_units           = 1
chromosomes_display_default = yes
...

creating the data

The data were created by a short script which generated 27 random data sets. Each data set had between 4 and 10 data points and for each data set the values were cumulative.

The data files are named species.0.txt ... species.26.txt.

# species.0.txt
gh 0 401 id=gh6
gh 401 468 id=gh8
gh 468 674 id=gh9
gh 674 1000 id=gh10

...

# species.26.txt
gh 0 235 id=gh3
gh 235 454 id=gh4
gh 454 534 id=gh7
gh 534 710 id=gh8
gh 710 757 id=gh9
gh 757 1000 id=gh10

Each data value has a unique id (gh1 ... gh10) which is used to assign a color.

This kind of data set corresponds to a situation where you have several samples (e.g. species) each composed of one or more components from a fixed set (gh1 ... gh10). The components are displayed as cumulative fractions and therefore each data set has the same maximum value.

automating track display

Let's display the first 7 data sets. We'll use automated track placement to do so, so that the same configuration file can be used for each track.

# circos.conf
<plots>
<<include speciesplot.conf>>
<<include speciesplot.conf>>
<<include speciesplot.conf>>
<<include speciesplot.conf>>
<<include speciesplot.conf>>
<<include speciesplot.conf>>
<<include speciesplot.conf>>
</plots>

where

# speciesplot.conf
<plot>
type  = highlight
file  = species.counter(plot).txt
r0    = eval(sprintf("%fr",conf(track_start) - conf(track_step) * counter(plot) ))
r1    = eval(sprintf("%fr",conf(track_start) - conf(track_step) * counter(plot) + conf(track_width) ))
fill_color       = black
stroke_thickness = 10p
stroke_color     = white
#stroke_color     = black
<<include ghcolorrule.conf>>
</plot>

You can futher simplify this configuration by moving the common constant parameters (e.g. fill_color) to the outer <plots> block. Any parameters in this block will be inherited by each of the <plot> blocks.

coloring data values

Data values are colored based on rules, which use the value of the id parameter. The rules are imported into each track block from ghcolorrule.conf which contains a single rule

<rules>
<rule>
condition  = 1
# ghN -> spectral-10-div-N
fill_color = eval(sprintf("spectral-10-div-%d",substr(var(id),2)))
</rule>
</rules>

The condition is always true, so the rule applies to each point. The number from the point's id field (e.g. gh7) is extracted and then used to define a color (e.g. spectral-10-div-7). The color palette used is the 10-color diverging spectral Brewer Palette.

Track Placement

Tracks are placed automatically, controlled by three track_* parameters

# if using 7 data sets
# well spaced
#track_width = 0.08
#track_step  = 0.1
#track_start = 0.9

# abutting
track_width = 0.1
track_step  = 0.1
track_start = 0.9

# if using 3x7 data sets
#track_width = 0.03
#track_step  = 0.04
#track_start = 0.95

To add more tracks, add more

<<include speciesplot.conf>>

lines to the <plot> block in circos.conf.