Hello,
I'm using CNVkit with default parameters and segementation with CBS.
I used a sample that has a duplication about 850 bin(17Mb) in chr5
However, the result didn't got the duplication, and I have another sample(duplication, 51 bins, 1.1Mb in chr22) , which also has the same problem.
Then, I checked the cnr data of CNVkit,and I found there were two bins with log2 value ( -20) far from the true duplication.
I tried:
- remove the low log2 value(-20), I got the right duplication
- changed the log2 value from -20 to -5, the duplication could be detected.
- changed the log2 value of neighbours to -20 (about 3 continuous bins), the duplication could also be detected.
I also tried these on another sample(duplication, 51 bins, 1.1Mb) with one low log2 ratio bin , the true duplication can be detected.
CNVkit set seed for CBS, and I tried remove the seed, and run CBS again. The duplication can be detected 10 out of 20 times. But, I tried to run 20 times on another sample(duplication, 51 bins, 1.1Mb), I didn't got the expected result.
I am wondering why the true CNV can be affected by the bin with low log2 ratio, maybe it's related to the algorithm of CBS. I tried the smooth-cbs suggested in cnvkit 0.9.7
, but it didn't work on the sample (duplication, 51 bins, 1.1Mb). The cnr file of the sample with a duplication in chr22 is attached.
10243.chr22.zip
Is there any suggestion on how to deal with this problem?
Thank you!
Hello,
I'm using CNVkit with default parameters and segementation with CBS.
I used a sample that has a duplication about 850 bin(17Mb) in chr5
However, the result didn't got the duplication, and I have another sample(duplication, 51 bins, 1.1Mb in chr22) , which also has the same problem.
Then, I checked the cnr data of CNVkit,and I found there were two bins with log2 value ( -20) far from the true duplication.
I tried:
I also tried these on another sample(duplication, 51 bins, 1.1Mb) with one low log2 ratio bin , the true duplication can be detected.
CNVkit set seed for CBS, and I tried remove the seed, and run CBS again. The duplication can be detected 10 out of 20 times. But, I tried to run 20 times on another sample(duplication, 51 bins, 1.1Mb), I didn't got the expected result.
I am wondering why the true CNV can be affected by the bin with low log2 ratio, maybe it's related to the algorithm of CBS. I tried the smooth-cbs suggested in cnvkit 0.9.7
, but it didn't work on the sample (duplication, 51 bins, 1.1Mb). The cnr file of the sample with a duplication in chr22 is attached.
10243.chr22.zip
Is there any suggestion on how to deal with this problem?
Thank you!