CNV calling with Control-FREEC? Significant calls

Question

[ivivek_ngs]

Dear All,

I have some high coverage exome data for my normal/tumor pairs where I am particularly interested in calling the CNV/ I am also having reprogrammed clones from the tumor to give IPSC(induced pluripotent stem cells). I am now trying to understand by CNV calling by the Control-FREEC method how the loss or gain in copy is conserved across tumor and its IPSC, does the genetic background is entirely wiped out still it is conserved and shows potential oncogenes. I have found some hits infact. The thing which bothers me is in Control-FREEC there is no option to give find how significant are the CNV calls, like you get an output file with .CNV which is having the coordinates with the copies gained or lost for each chromosomes but how can we select which should be the true positives. I have 90 CNVs called for my tumor and around 200 for my IPSC, out of them barely 50% coincides with the known CNVs when I tried to merge them with the CNV database making them TP but then rest can either be FP or novel as well. Is there any way to judge that with CNV calls from Control-FREEC. The software is excellent with good visualization plot but I am confused in this part . Would be happy if someone shares some ideas. Also I would like to know is it possible to have 8 or 9 copies for a region on the chr , the same has been mimicked in my IPSC as well for the exact region but is it feasible to have such high copy number ? I would be grateful if someone can share some light.

[ivivek_ngs]

I did some further analysis with the Control-FREEC tool. I am not aware of the ploidy status of my samples so the first analysis I did on my tumor and clones were keeping the ploidy status as 2 and retrieved the CNVs and now I tested for with ploidy status 3 and I get 50% different CNV calls. The supplementary paper says if you do not have the ploidy information try to use the different ploidy states to get the best CNV calls but then how will I asses for the true positive? how I will get to know which ploidy status gives much more accurate call? Can anyone give me some suggestions?

[mittu1602]

Hello, I would like to know if any relevant answers was obtained for this issue. because I am facing the same type of issue for CNV analysis.

Answer 1

[ivivek_ngs]

I should have updated this way ago when I found out a way how to perform the significant calls. As you may know, Control-FREEC gives you a .CNV file which can then we run independently on a script in R that the group provides and test the significance of those calls. Find it here. One piece of advice read the manual properly.

Code snippet from the manual

Calculate significance of Control-FREEC predictions with an R script

One can add the p-values to predicted CNVs by running assess_significance.R:
cat assess_significance.R | R --slave --args < *_CNVs > < *_ratio.txt >
R and package rtracklayer should be installed!
Running this script will add both Wilcoxon test and Kolmogorov-Smirnov test p-values to each line of _CNVs file.
It will also add headers to the columns of _CNVs.

For the ploidy status you will have to use other tools to estimate the ploidy otherwise use the default. Again read the FREEC versioning well and v8 can now estimate the average ploidy status.

Starting from version v8.0, we provide a possibility to detect subclonal gains and losses and evaluate the likeliest average ploidy of the sample. Also, the procedure for evaluation of tumor purity has been improved.

Good luck!

[mittu1602]

Thank you for taking out time and updating it. I hope it helps me.

[ivivek_ngs]

if you find the answer useful please upvote or accept it as an answer to make the thread complete for others to use it for future reference.