Use the latest version of Circos and read Circos best practices—these list recent important changes and identify sources of common problems.
If you are having trouble, post your issue to the Circos Google Group and include all files and detailed error logs. Please do not email me directly unless it is urgent—you are much more likely to receive a timely reply from the group.
Don't know what question to ask? Read Points of View: Visualizing Biological Data by Bang Wong, myself and invited authors from the Points of View series.
If you are having trouble with installation of Perl or modules, use online resources that explain the details of how to download Perl, get it working (Linux, Mac OS X, Windows [win32.perl.org wiki, ActiveState, Strawberry]), and how to install modules (UNIX, Windows). If you're still stuck, post your questions to the Circos group.
Having trouble with libgd and GD? See the Perl Monks libgd/GD Tutorial, Shaun Jackman's Homebrew formula, Wang's install zlib/libpng/jpeg/freetype/libgd/GD on Mavericks as well as my own guide for installation of libpng, freetype, libgd and GD on Mac OS X Mavericks. There are some useful threads in the Google Group about this.
Stumped by an error? A good strategy is to Google the error message (e.g. mkdir /usr/local/share/man: permission denied) to find the solution.
Want to learn more about Perl? Try learn.perl.org.
Circos generates static images. The image generation process is driven by a central configuration file. This file usually imports other configuration files, such as global color and font settings.
There is no interface to Circos. Workflow typically proceeds as follows
Configuration files are parsed using Config::General module. All pertinent features are described below, but for those so inclined, refer to this module's man page to learn about the details of the syntax and parsing of these files.
Settings are defined in configuration files using the format
variable = value
Note that although Config::General
supports a whitespace as an assignment delimiter, Circos requires that
= for all definitions.
Some settings are grouped in blocks, to create a hierarchical structure.
<ideogram> thickness = 30p fill = yes ... </ideogram>
Some blocks can have multiple instances, such as data tracks. Typically, these are enlosed in another block, here <links>.
<links> <link> file = data/set1.txt color = black ... </link> <link> file = data/set2.txt color = red ... </link> </links>
Please ensure that all your configuration blocks are correctly terminated with an appropriate close block tag.
<ideogram> ... <ideogram> # <-- if this is missing, an error will result
There are four places in which a configuration parameter can be specified.
In order of increasing importance
A parameter set by a rule overrides any specified in the data file. A data file parameter overrides any in a <plot> or <link> block, which in turn override any global parameters in a <plots> or <links> block.
<plots> # global parameter fill_color = white <plot> ... # other plot parameters, such as file, type, position, etc # local to this track - overrides the fill_color=white value fill_color = grey # in data.txt: # # ... # hs1 10 20 0.50 fill_color=dblue # ... # # The data point will be purple, overriding both fill_color=red and fill_color=green values. <rules> <rule> condition = var(value) < 0.33 # specific to data points matching the condition, overrides any previously specified # fill_color value, (global, local, data file) fill_color = orange </rule> </rules> </plot> </plots>
If you are plotting a large number of similar tracks (e.g. groups of histograms, heat maps, etc), it is useful to apply global parameters where possible. This is particularly effective when combined with automated track placement with track counters. In the example below, each plot will be a heatmap with the same min/max value and color map. Within individual <plot> blocks, only the parameters specific to that block need to be specified.
<plots> type = heatmap min = 0 max = 1 # this color name is a list - see below about color lists color = spectral-4-div <plot> file = data.1.txt r1 = 0.6r r0 = 0.5r ... </plot> <plot> file = data.2.txt r1 = 0.7r r0 = 0.6r ... </plot> <plot> file = data.3.txt r1 = 0.8r r0 = 0.7r ... </plot> <plot> file = data.4.txt r1 = 0.9r r0 = 0.8r ... </plot> … </plots>
Some settings never or rarely change, such as colors and fonts. To keep the main configuration file modular, the files for these static values are imported using the <<include ...>> directive.
Two files should always be imported from
etc/ in the
Circos distribution. These are
# colors, fonts and fill patterns <<include etc/colors_fonts_patterns.conf>> # system and debug parameters <<include etc/housekeeping.conf>>
etc/colors_fonts_patterns.conf file itself imports several files from the Circos distribution.
# etc/colors_fonts_patterns.conf <colors> import(etc/colors.conf) </colors> <fonts> import(etc/fonts.conf) </fonts> <patterns> import(etc/patterns.conf) </patterns>
Circos requires that these blocks be present and populated with definitions.
Conventionally, I store configuration for ideograms in an external
ideogram.conf) and for tick mark formatting (
reason for this is that these settings are fairly verbose, but are not
related to a data set. By importing ideogram and tick mark settings
from an external file
<<include ideogram.conf>> <<include ticks.conf>>
the main configuration file is kept more succinct. Moreover, if you are creating multiple images with different data sets, you are likely to use the same settings for ideogram and tick mark formats. Storing these settings separately just makes sense.
are placed in the same directory as
circos.conf, so no
path (relative or absolute) needs to be added to the filename in the
<<include >> directive.
If you have parameters that rarely change, consider creating external configuration files for these.
You can use the <<include >> directive anywhere in the configuration file, such as in plot blocks.
<plot> file = data.4.txt r0 = 0.8r r1 = 0.9r <<include plotsettings.conf>> </plot>
Inclusion can be arbitrarily nested. In other words, included files can themselves include others, and so on.
When you include a file using <<include CONFIG_FILE_PATH/CONFIG_FILE>>, Circos will search the following paths for the file
In the configuration file, parameters are typically set to constants using the syntax
variable = value
color = blue
There will be times when you'll want to specify the value of a configuration parameter using the value or a function of another.
Any parameter can be set to the value of another parameter using the syntax
parameter2 = conf(parameter1)
or, for parameters which are found in blocks
parameter2 = conf(block1,parameter1) parameter2 = conf(block1,block2,parameter1) ...
track_color = blue <plots> <plot> color = conf(track_color) ...
When the configuration file is parsed, simple substitution is
exhaustively made until all
have been replaced by values.
Make sure that you include the full block path of the parameter when using this syntax. Thus for
<block1> <block2> parameter1 = ... </block2> </block1>
you would use
Any parameter can be written as Perl code and evaluated at
run-time. To use this feature, enclose the parameter in an
thickness = eval(1+1) color = eval("b"."l"."u"."e")
eval() feature is very useful when used to refer to and manipulate other configuration parameters.
track_color = blue track_width = 100 track_start = 0.5 <plots> <plot> # color=blue color = conf(track_color) # r0 = 0.5r r0 = eval(conf(track_start) . "r") # r1 = 0.5r+100p r1 = eval(conf(track_start) . "r" + conf(track_width) . "p") </plot> </plots>
If you are defining a parameter by a single
value, you do not need
eval(), since only a substitution is
required. However, if you need to manipulate this value (e.g. append a
string, perform arithmetic), then
eval() is required
since the expression must be evaluated as code.
eval()calls require that you write correct
Perl syntax, the chance of a mistake and therefore fatal error is
high. Double check! It is common to forget to quote text in these
calls — verify that you are not using a value as a bare word.
## OK x = eval( 1.05 . "r" ) ## NOT OK - r is meant to be a string, but without quotes Perl will ## interpret it as a bare word, producing an error x = eval( 1.05 . r )
eval() in <rule> blocks are evaluated independently for each data point.
Circos keeps a running count of the tracks as they are drawn. You can use these variables to fully automate track placement.
See Automating Tracks.
By including the
etc/colors_fonts_patterns.conf file in the main configuration file
# circos.conf <<include etc/colors_fonts_patterns.conf>> ...
you are including definitions for primary RGB and HSV colors. Also defined are Brewer palette colors and the conventional human chromosome color palette. To learn more about Brewer palettes, see my Color Palettes Matter presentation.
etc/colors.conf file, which is included by
etc/colors_fonts_patterns.conf, itself includes these various color definitions
# etc/colors.conf # primary RGB colors ... # Brewer palettes # see etc/colors.brewer.conf <<include colors.brewer.conf>> # UCSC genome browser human chromosome colors # see etc/colors.ucsc.conf <<include colors.ucsc.conf>> # HSV pure colors # see etc/colors.hsv.conf <<include colors.hsv.conf>>
Colors are referenced using their RGB values or their names (see below).
# using RGB values color = 107,174,241 # using name color = blue
When passing a color as an option in data files, the RGB values need to be delimited by
(...). For example, if you want to add a color to a link
# using a color name chr1 100 200 chr2 200 250 color=blue,thickness=2 # using RGB value chr1 100 200 chr2 200 250 color=(107,174,241),thickness=2
Colors in Circos are defined by their RGB or HSV values and specified by a name (e.g. red, orange, etc). Many named colors are pointers to Brewer palette equivalents.
# pure orange porange = 255,127,0 # dark pure orange dporange = 234,110,0 # points to Brewer color... orange = oranges-7-seq-4 # ...which is defined in colors.brewer.conf as oranges-7-seq-4 = 253,141,60
Typically for a given color root name (e.g. orange), there are
corresponding shades of the color with prefixed
l (light). The light version may be prefixed by one or more
(very). These shades point to a sequential Brewer palette for the color. For example, oranges point to the 7-color 'oranges' Brewer palette
vvlorange = oranges-7-seq-1 vlorange = oranges-7-seq-2 lorange = oranges-7-seq-3 orange = oranges-7-seq-4 dorange = oranges-7-seq-5 vdorange = oranges-7-seq-6 vvdorange = oranges-7-seq-7
which is defined in
oranges-7-seq-1 = 254,237,222 oranges-7-seq-2 = 253,208,162 oranges-7-seq-3 = 253,174,107 oranges-7-seq-4 = 253,141,60 oranges-7-seq-5 = 241,105,19 oranges-7-seq-6 = 217,72,1 oranges-7-seq-7 = 140,45,4
If you want the pure, saturated version of the color, use the
p prefix. For example,
porange is a pure bright orange.
vvlporange = 255,182,106 vlporange = 255,164,82 lporange = 255,146,54 porange = 255,127,0 dporange = 234,110,0 vdporange = 213,92,0 vvdporange = 193,75,0
etc/colors for the full list of colors.
I suggest that you try using the Brewer colors (e.g.
porange), because they are perceptually uniform. However,
they will appear less punchy and saturated than their pure
equivalents. In particular, the Brewer reds may appear pinkish and
light when used on their own.
Experiment, but be aware of the perceptual aspects of color, which will influence how your figure is perceived (see my Color Palettes Matter presentation).
Brewer colors are categorized into one of three palette types: sequential, diverging and qualitative. For a given palette type (e.g. sequential), there are a variety of palettes (e.g. reds, greens, blues). Each palette is available for various number of colors (e.g. 3, 4, 5, ...).
The syntax for a Brewer color name is
palettename-ncolors-palettetype-index. The palette names, for each type, are
# sequential (-seq-) (3-9 colors) blues bugn bupu gnbu greens greys oranges orrd pubu pubugn purd purples rdpu reds ylgn ylgnbu ylorbr ylorrd # diverging (-div-) (3-11 colors) brbg piyg prgn puor rdbu rdgy rdylbu rdylgn spectral # qualitative (-qual-) (3-8 colors, some up to 12 colors) accent (3-8 colors) dark2 (3-8 colors) paired (3-12 colors) pastel1 (3-9 colors) pastel2 (3-8 colors) set1 (3-9 colors) set2 (3-8 colors) set3 (3-12 colors)
For example, purple-orange diverging 9-color palette colors are
You can use the HSV color space to define colors. To do so, specify the HSV values as hsv(h,s,v). For example,
red = hsv(0,1,1)
All pure HSV colors (s = 1, v = 1) are defined in
hue000 = hsv(0,1,1) hue001 = hsv(1,1,1) ... hue359 = hsv(359,1,1) hue360 = hsv(360,1,1) # same as hue000
etc/colors.unixnames.txt defines a large
number (700+) of named colors, taken from UNIX's rgb.txt
file. This file is not included by default.
Many definitions in
this file duplicate definitions in
colors.unixnames.txt defines blue as 0,0,255 but in
colors.conf it is blues-7-seq-4, which is 107,174,214). Including
colors.unixnames.txt together with
(colors.conf) will result in an error.
You can assign an alpha channel value to a color (transparency) by including a fourth component.
# 0 < alpha < 1 # 0 opaque # 1 transparent red_faint = 255,255,255,0.8 # or alpha 0-127 # 0 opaque # 127 transparent red_also_faint = 255,255,255,102
You can use either the [0,1] range for the alpha value, or [0,127]. In both cases, the right end of the interval corresponds to transparent. For example, if alpha is in the range 0-127 then
a=0 corresponds to fully opaque, and
a=127 to fully transparent.
Please see Transparent Link tutorial for discussion about automating definition of these colors.
To create a fully transparent color (e.g. for an image with transparent background), you'll need to define a color named
transparent. A transparent color still requires an RGB value (a strange artefact in gd implementation). Choose an RGB value that you aren't using elsewhere. Typically something like 1,0,0 will be suitable.
# in color.conf transparent = 1,0,0
The transparent color will be available using the name
transparent. A synonym
clear is also provided. To use the transparent color (e.g. for background),
<image> ... background = transparent # 'clear' also works here ... </image>
clear are reserved. Do not use these two color names for other colors.
You can include synonyms for colors, by defining one color using the name of another color, instead of RGB or RGBA values.
favourite = green almost_favourite = orange ... green = 51,204,94 orange = 255,136,0
Be careful not to create infinite lookup loops — these produce an error.
# don't do this favourite = green green = favourite
A color list can be defined by specifying a comma-delimited list of existing colors
red_list = dred,red,lred,vlred
or, more conveniently, a regular expression. The results will be sorted by the value of any capture buffers. The order will be reasonable (numerically or alphanumerically depending on the value of the capture buffer). If you want to sort the matches in reverse, wrap the regular expression in
For example, to create a list of the 9-color spectral Brewer palette,
spectral9 = spectral-9-div-(\d+)
and to create a reversed list
spectral9r = rev(spectral-9-div-(\d+))
Color lists are used with heat maps.
Lists for all Brewer palettes are predefined (see
etc/brewer.lists.conf). For a given color set
name-ncolors-type-index, two lists are available
name-ncolors-typeBrewer palette color list (e.g.
reds-8-seq = reds-8-seq-1,reds-8-seq-2,...)
name-ncolors-type-revcorresponding palette, with colors in reverse order (e.g.
reds-8-seq-rev = reds-8-seq-8,reds-8-seq-7,...)
For example, the 6-color Brewer palette lists that are defined are
# sequential blues-6-seq bugn-6-seq bupu-6-seq gnbu-6-seq greens-6-seq greys-6-seq oranges-6-seq orrd-6-seq pubu-6-seq pubugn-6-seq purd-6-seq purples-6-seq rdpu-6-seq reds-6-seq ylgn-6-seq ylgnbu-6-seq ylorbr-6-seq ylorrd-6-seq # diverging brbg-6-div piyg-6-div prgn-6-div puor-6-div rdbu-6-div rdgy-6-div rdylbu-6-div rdylgn-6-div spectral-6-div # qualitative accent-6-qual dark2-6-qual paired-6-qual pastel1-6-qual pastel2-6-qual set1-6-qual set2-6-qual set3-6-qual
Each has a
-rev (reversed) counterpart (e.g.
These lists are automatically imported from
etc/colors.brewer.conf. Thus, if you import the Brewer colors (done by default), you are automatically including all Brewer lists.
Brewer palettes provide sets of perceptually uniform colors and should be used whenever possible (i.e., always).
Additionally, color sets of pure HSV colors (s = 1, v = 1) are defined in
colors.hsv.conf. Two kind of HSVG color lists are defined.
Lists of colors by hue step are named hue-sH, for a set of colors that vary by a change in hue of H. For example,
hue-s45 includes the colors
hue315. Lists for steps 2, 3, 4, 5, 6, 8, 9, 10, 12, 15, 18, 20, 24, 30, 36, 40, 45, 60, 72, 90, 120, 180 and 360 are defined.
The other set of lists are named hue-N, for a set of N uniformly spaced colors. For example,
hue-7 includes the colors
hue309. Lists for 3 to 30 colors are defined.
Generating the color lists can take several seconds. For this reason, Circos employs a caching mechanism to store color lists definitions. By default, the cache file is
/tmp/circos.colorlist.dat. If the cache is older than the configuration file, or color definitions, it is recomputed. The length of time required is a function of the total number of colors (color definitions multiplied by automatic transparency levels) and the number of lists. If you are trying to optimize image generation speed, and do not wish to count on caching, remove any list definitions you are not using and reduce the number of automatic transparency levels.
A set of colors named after chromosomes is also defined and corresponds to the chromosome color scheme used by UCSC Genome Browser and other online resources. This is a standardized palette.
chr1 = 153,102,0 chr2 = 102,102,0 chr3 = 153,153,30 ... chrX = 153,153,153 chrY = 204,204,204
Another set of colors is named after cytogenetic band colors, typically reported in karyotype files. These colors define the G-staining shades seen in ideograms.
gpos100 = 0,0,0 gpos = 0,0,0 gpos75 = 130,130,130 gpos66 = 160,160,160 gpos50 = 200,200,200 gpos33 = 210,210,210 gpos25 = 200,200,200 gvar = 220,220,220 gneg = 255,255,255 acen = 217,47,39 stalk = 100,127,164
I strongly suggest that you place new color definitions in a separate file. Modularity will make maintenance easier. And given that you'll likely want access to your custom colors for all images, include them globally rather than on an image-by-image basis.
For example, if you create your own blue
# in mycolors.conf niceblue = 17,111,227
you can include this file like this
# all default color definitions <<include colors_fonts_patterns.conf>> # this will append your definitions to the <colors> block <colors> <<include mycolors.conf>> </colors>
You can quickly add colors directly
# all default color definitions <<include colors_fonts_patterns.conf>> # this will append your definitions to the <colors> block <colors> <<include mycolors.conf>> niceblue2 = 37,101,179 </colors>
Circos uses CMU Modern fonts. These are found in
fonts/ in the distribution and are associated with unique definitions
(e.g. light, bold, italic) used in configuration files.
light = fonts/modern/cmunbmr.otf # CMUBright-Roman normal = fonts/modern/cmunbmr.otf # CMUBright-Roman default = fonts/modern/cmunbmr.otf # CMUBright-Roman semibold = fonts/modern/cmunbsr.otf # CMUBright-Semibold bold = fonts/modern/cmunbbx.otf # CMUBright-Bold italic = fonts/modern/cmunbmo.otf # CMUBright-Oblique bolditalic = fonts/modern/cmunbxo.otf # CMUBright-BoldOblique italicbold = fonts/modern/cmunbxo.otf # CMUBright-BoldOblique
To use a specific font for an element, specify its label (e.g. normal, bold) in the configuration file. The default fonts are shown here.
To add your own fonts, copy the TTF file to
add a new label to the font in the
To render very small labels, you should consider using bitmapped fonts. These fonts are designed to be used at a specific size and without anti-aliasing. One such family is the Mini Font set from minifonts.com.
Other legible fonts, commonly used in terminals and text editors are
These fonts are shown in the image section of this tutorial.
If you suspect there may be a problem with drawing images, please run
and look at the output
gddiag.png. It should look like the image in this tutorial.
Many quantities defined in the configuration files require units, which are one of
Unit designations are suffixed to the value and may be mixed
# 1 pixel padding padding = 1p # relative padding (e.g. relative to label width) padding = -0.25r # radius of track (relative to inner ideogram radius) r0 = 0.5r # combination of relative and pixel values r1 = 0.5r+200p
The reason why Circos insists on units is to reduce the strain of interpreting the configuration file parameters - a large number of custom values can quickly make the file opaque to quick inspection.