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 from the Points of View series.
Copy number values, such as those collected from microarray or clone-based hybridization experiments, are typically drawn as scatter plots. In this tutorial, I'll show how CNV values can be drawn with Circos.
We'll assume that the CNV data exists for various points in the genome, formatted for a scatter plot
... hs1 30200000 32400000 0.798258 hs1 32400000 34600000 0.0495811 hs1 34600000 40100000 -2.46056 hs1 40100000 44100000 1.40846 hs1 44100000 46800000 -2.35913 ...
If you are using very dense microarrays, the number of individual values can exceed 100,000. In this case, you should window the data to reduce the number of points.
Individual probes may have single base positions, such as
hs1 3021532 3021532 0.798258
but when using clone-based hybridization, or a windowed data set, a single CNV value is associated with a range
hs1 30200000 32400000 0.798258
In this case, keep the input data intervals as a range. Although a scatter plot glyph is always placed in the center of the range (and therefore the output is indistinguishable from a data set in which the range is the midpoint), if you choose to draw the data as a histogram, bin width will be based on the range size. In addition, you have access to the range sizes in the rules block, if you need them.
For this tutorial, I've generated random CNV values, in the range [-3,3].
We'll want to have a different background for the positive and negative CNV values. To do so, you'll need to define two separate scatter plot tracks. You'll use the same input data for both, and distinguish them by different min/max values.
The track for negative data will be positioned at 60-75% of the circle's radius and show data in the range
<plot> type = scatter file = data/8/13/data.cnv.txt r0 = 0.6r r1 = 0.75r min = -3 max = 0 glyph = circle glyph_size = 8 fill_color = red axis = yes axis_color = lred axis_thickness = 2 axis_spacing = 0.25 background = yes background_color = vlred_a8 ...
The track for positive values is defined analogously, except that
r0, r1, min, max are different, as are the axis and background colors.
<plot> type = scatter file = data/8/13/data.cnv.txt r0 = 0.75r r1 = 0.9r min = 0 max = 3 glyph = circle glyph_size = 8 fill_color = green axis = yes axis_color = lgreen axis_thickness = 2 axis_spacing = 0.25 background = yes background_color = vlgreen_a8 ...
Rules are used to dynamically adjust features of the figure based on data values. In this example, I'll modify both the size and the outline of the glyphs. The size will be proportional to the absolute value, and for extremely large (or small) values, we'll put a black outline around the glyphs.
<rules> # change the glyph_size for each point using formula 6+4*abs(cnv) <rule> importance = 100 condition = 1 glyph_size = eval( 6 + 4*abs(_VALUE_)) flow = continue </rule> # if the value is >2, add a black outline <rule> importance = 90 condition = _VALUE_ > 2 stroke_color = black stroke_thickness = 2 </rule> </rules>
You may wish to indicate which parts of the genome contain positive (or negative) CNV values. Because the points within the scatter plot are, well, scattered, it can be hard to immediately identify regions of positive or negative CNV.
One way to achieve this using a heat map, which is covered in the another tutorial. You can use the same input data for the heat map track.
However, here I'll show how to create a similar track by collapsing the scatter plot values onto a single radial position. By definingtwo more scatter plots, similar to those above, but with
r0=r1, the points will be drawn at the same radial position, regardless of their value. Combining this with a rule that changes the transparency of the color of the point based on the value, we get something that looks like a heat map.
<plot> type = scatter file = data/8/13/data.cnv.txt r0 = 0.955r r1 = 0.955r min = 0 max = 3 glyph = square glyph_size = 8 fill_color = green <rules> <rule> importance = 100 condition = 1 fill_color = eval( "green_a" . min(10,max(1,int((3-abs(_VALUE_))/0.30)))) </rule> </rules> </plot>
You should use this approach for cases when a heat map is insufficient. For example, you can use circular glyphs and map their size by value to create a string-of-pearls effect.