Need Help Understanding Segdup File Format And Generating Circos Plots

Question

[Jordan]

Hi,

I have used SVDetect to generate links. And I got it converted to circos input format (segdup.txt). I don't understand the logic of this format. Here are first few lines:

1    hschr1    10000005    10000080    color=white
1    hschr18    54254700    54254735    color=white
2    hschr1    10000008    10000083    color=white
2    hschr18    19966691    19966726    color=white
3    hschr1    10000016    10000051    color=white
3    hschr16    66915746    66915821    color=white
4    hschr1    10000016    10000051    color=grey
4    hschr17    29951284    29951359    color=grey
5    hschr1    10000016    10000051    color=white
5    hschrX    15854763    15854838    color=white

Why is the color important? I understand it has something to do with generating circos plots. Nothing more.

And without much knowledge, I tried generating circos plots anyways, to see how the output looks like. Surprisingly there are no links in it.

SVDetect gave an example with the old version of circos plots. The circos conf format has changed quite a bit with the new one and I have modified everything to work in the current format. Just when I thought everything works fine, there are no links generated.

Here is a part of the (now updated it to the entire link section) link section in circos conf file (after my modification looking at the tutorial):

<link>
show         = yes
color        = vvdgrey
thickness    = 2
file         = sample.ab.bam.links.compared.segdup.txt
#record_limit = 2500

<rules>

<rule>
importance = 10
condition  = var(color) eq "red"
thickness  = 3
z          = 10
</rule>

<rule>
importance = 9
condition  = var(color) eq "orange"
thickness  = 3
z          = 9
</rule>
<rule>
importance = 8
condition  = var(color) eq "purple"
thickness  = 3
z          = 8
</rule>

<rule>
importance = 7
condition  = var(color) eq "green"
thickness  = 3
z          = 7
</rule>

<rule>
importance = 6
condition  = var(color) eq "blue"
thickness  = 3
z          = 6
</rule>

<rule>
importance = 5
condition  = var(color) eq "black"
thickness  = 3
z          = 5
</rule>
85
<rule>
#show         = no
importance = 4
condition  = var(color) eq "grey"
thickness  = 2
z          = 4
</rule>
</rules>
</link>
</links>

I only modified the condition in original conf file provided by SVDetect, which looked like this:

condition = _COLOR1_ eq "red"

Of course I have tried using the original conf file as well, but still it gave me the same image.

To summarize, if anyone can explain the format of the segdup file and how to generate links properly using circos plot I would really appreciate it.

Thanks!

Answer 1

[SES]

It looks like your conf file is not formatted correctly. You need to have a section that defines the links with a <links> tag and for each link you need a closing link tag. Also, you are missing the closing links tag above. From the examples,

<links>
show          = __$CONF{show_links}__
ribbon        = yes
flat          = yes
radius        = 0.45r-55p  
bezier_radius = 0r
color         = black_a5

<link ratbundle>
file = exampledata/rn.bundle.txt
<<include link.rules.conf>>
</link>

<link mousebundle>
file = exampledata/mm.bundle.txt
<<include link.rules.conf>>
</link>

</links>

I would try to get this example to work first and then try with your data.

[Jordan]

The link section is only a part of it, like I said above. I didn't display the closing links here. I have tried to produce example links from the tutorial and they did work. It's only for this that I'm having the trouble.

When you say formatting is not right, can you tell me what you are talking about?

[SES]

I am referring to the tags above and below the links section. Notice that you do not have an opening and closing link tag for the link in your example and you do not have a closing links tag. By tag, I am referring to the angled brackets. I don't see how it could work as you've formatted it above.

[Jordan]

Ok. So I updated the question with the entire links section. Since it was taking too much space I didn't want to put the entire block here.

[SES]

I still don't think it is formatted correctly. Notice in my example there is a links tag that defines the link attributes and then there are individual link file/confs below that. Each link has an opening and closing tag and there is an opening and closing tag for the whole links section. An easy test would be to try the example formatted as it is, and then remove the tags (as in your conf file) and see what the result is. Because they are not formatted correctly, I think you'll find the links won't work when changing the format, but as I said, it is easy to test.

[Jordan]

I took your advice and ran the circos plot on the link file you gave me. And it still did not give me any links. Here is my entire conf file:

<<include colors_fonts_patterns.conf>>
<<include ideogram.conf>>
<<include ticks.conf>>
<<include housekeeping.conf>>

# specify the karyotype file here; try also
karyotype = karyotype.human.hg19.txt

<image>
<<include image.conf>>
</image>

#chromosomes_units           = 1000000

chromosomes_units = 1000000
chromosomes_display_default = yes

<links>
show          = __$CONF{show_links}__
ribbon        = yes
flat          = yes
radius        = 0.45r-55p  
bezier_radius = 0r
color         = black_a5

<link ratbundle>
file = data/rn.bundle.txt
<<include etc/link.rules.conf>>
</link>

<link mousebundle>
file = data/mm.bundle.txt
<<include etc/link.rules.conf>>
</link>

</links>

And here is the runtime debugging:

debuggroup summary 0.17s welcome to circos v0.64 2 May 2013
debuggroup summary 0.17s loading configuration from file sample.circos.conf
debuggroup summary 0.17s found conf file sample.circos.conf
Use of uninitialized value $target in substitution (s///) at /data/sam/circos/0.64/bin/../lib/Circos/Configuration.pm line 606.
debuggroup summary 0.51s debug will appear for these features: summary
debuggroup summary 0.51s parsing karyotype and organizing ideograms
debuggroup summary 0.72s applying global and local scaling
debuggroup summary 0.73s allocating image, colors and brushes
debuggroup summary 8.23s drawing highlights and ideograms
debuggroup summary 8.84s found conf file /data/sam/circos/0.64/bin/../etc/tracks/link.conf
debuggroup summary,output 8.86s generating output
debuggroup summary,output 9.63s created PNG image ./circos.png (272 kb)
debuggroup summary,output 9.63s created SVG image ./circos.svg (120 kb)

Now I'm not sure what the problem is.

[SES]

The conf file I posted was just an example, I wouldn't expect it to work with your data. Can you try the examples in the circos /examples folder and confirm you can get them to work? That is what I was referring to about testing the conf format, sorry if that was unclear.

[Jordan]

Yes. It did work for me. And also tried the run script they had. I gave me a circos plot with links, histogram and heatmaps I think.

Answer 2

[Charles Warden]

I personally had a hard time getting SVDetect to produce the circos plots, and I found using POMO to plot the variants listed in the results table to be way easier:

http://pomo.cs.tut.fi/

If you want to visualize rearrangemnets as well as events like duplications, deletions, etc., you'll need to create two imput files (an association file for the stuff you want to form lines crossing the circle and an annotation file for events that should be specified with a single genomic coordinate). However, I still found it easier to reformat these two input files than try to get circos working properly with the intented files from SVDetect.

POMO Input File Format: http://code.google.com/p/pomo/wiki/AssociationAnnotationSyntax

[Jordan]

Hi.. I looked at the link. Looks very interesting. Even in this there is a color value in each row. Do you know what these colors mean by any chance?

[Charles Warden]

The POMO colors mean whatever you want them to mean. I could recommend manually specifying the color, so you can create your own legend (for duplications, deletions, etc.).