Scatter plot seems to have noise

etal / cnvkit

Copy number variant detection from targeted DNA sequencing

http://cnvkit.readthedocs.org

Other

540 stars 164 forks source link

Scatter plot seems to have noise #617

Open jeongmeani opened 3 years ago

jeongmeani commented 3 years ago

Hi,

Using cnvkit, i made cnv scatter plot. but the plot is weird like below. test

i expected the plot to be more clear in y-axis plot like below MSD-21C-04-B

why does this image have this noise?

Best regards

Jeongmin

tetedange13 commented 3 years ago

Hi @jeongmeani ,

Not an author of CNVkit, but as far as I know the CNVkit's scatter plot is a mere representation of the data contained in ".cnr" and/or ".cns" files (so are most CNVkit's nice plots)

To diagnose a possible issue, could you please give also details about:

Caracteristics of the dataset you are running CNVkit on (panel/WES/WGS? Amplicon/capture? DNA/RNA? Public/private dataset ...)
CNVkit's sub-commands you used to obtain such data (simply batch? Each pipeline step separately?) + exact parameters you used for each sub-command

Thanks. Have a nice day. Felix

jeongmeani commented 3 years ago

Hi, Felix.

sorry for late reply.

i let you know about information you request.

Caracteristics of the dataset i used is WES data.
yes, i ran using command batch and my sub-command is like below

1) cnvkit.py batch {Tumor-BQSR.bam} --normal {Normal-BQSR.bam} \ --targets $bed \ --fasta hg19.fa --access /data/tools/CNVkit_install/data/access-5k-mappable.hg19.bed \ --output-reference ref.cnn --output-dir $CNV_kit_dir 2) cnvkit.py scatter -s $CNV_kit_dir/$Sample_N'.BQSR.cn'{s,r} --y-max 10 --y-min -10 -o $CNV_kit_dir/$Sample_N'.png'

also, i made scatter plot of normal cnr and cns file reflecting normal.ref.cnn. but this plot too weird. i guess that this plot should be shown more clear like this? MSD-21C-04-C

but the plot is shown like below

can you recommend anything?

Best regards

Jeongmin

tetedange13 commented 3 years ago

Thanks @jeongmeani ,

So matched tumor/normal situation? Also which version of CNVkit are you on please? (sometimes simply upgrading to latest one could solve smthing) => Just for your information, in case you would have several pairs of samples like this, CNVkit's documentation advice to pool all your normal samples beforehand, then to run each tumor sample against your PoN
Also I do not see any --drop-low-coverage in your batch command? => In this other part of CNVkit's documentation, it is adviced to remove such low covered bins => This is my strongest hint, give it a try and tell us what you get?
Also do you have an idea of the mean coverage of your WES? Above 10X?

jeongmeani commented 3 years ago

Hi, Felix.

Sorry for late reply.

yes, this is tumor/normal matched situation. and CNVkit version is 0.9.6.dev0 on python3.5
as your recommend, i have used --drop-low-coverage option, but there is no change.
yes Above 10X, mean coverage of the data is above 100X

so, would you recommend anything of this?

Best Regards

Jeongmin

tetedange13 commented 3 years ago

Hi @jeongmeani ,

Sorry I run out of ideas... => Try to update to a newer version of CNVkit (v0.9.9) ? (+ try with/without --drop-low-coverage ?) => This could be linked to an issue fixed since your version => Or simply a known issue on WES datasets? Unfortunately I am not expert enough to tell..

Have a nice day. Kind regards. Felix.

tetedange13 commented 3 years ago

Hello @jeongmeani ,

Any news since last time ? => I do not know if @tskir saw this and can maybe have any idea?

Have a nice day. Kind regards. Felix.