CNANorm: A normalization method for Copy Number Aberration in cancer samples. - Page 2

christophe · 02-02-2012, 09:57 AM

Hi Stefano,
Thanks for the CNAnorm Package.

As PeakPloidy is really computationally time consuming, is there a way to run it as a parallel job (multithreading) using an R package such as "snow" or similar ?
The time to compute a 60MB ".tab" file is long, very long :-)

Thanks
Chris

stefanoberri · 02-02-2012, 02:53 PM

Quote:

Originally Posted by christophe

As PeakPloidy is really computationally time consuming, is there a way to run it as a parallel job (multithreading) using an R package such as "snow" or similar ?

No, unfortunately I haven't thought about parallel approach at the design stage. However, here a suggestion on how to make it fast and yet have high resolution.

From the same bam files, you produce two tab files (using bam2windows.pl) One at high resolution, one at low resolution (for instance with --window 100000). At this point you load and perform the usual steps using the low resolution table (let's say creating a CNlow object) up to the "validation" step. For the normalisation, high resolution is not very important. As long as you have enough windows (a few thousand) it will work.
Now you load the high resolution file (let's say creating CNhigh object) and perform usual GC correction, but do NOT perform peakPloidy (the computational intesive step).
You move to the discreteNorm step

Code:

CNhigh <- discreteNorm(CNhigh, normBy = CNlow)

Because it is the same sample, you can transform the high resolution data by the low resolution data. It might actually be better as there is less noise.

alternatively, especially for testing/debugging, you can use

Code:

CN <- peakPloidy(CN, exclude = toSkip, method = 'density')

this is largely independent of number of windows, so should be reasonable.

I hope this help

Amjad85 · 01-06-2013, 04:47 AM

Hi,

Is there a function to merge the segments (similar adjacent windows) in the output file resulted from exportTable to produce CNA calls?

stefanoberri · 01-07-2013, 01:14 AM

Quote:

Originally Posted by Amjad85

Hi,

Is there a function to merge the segments (similar adjacent windows) in the output file resulted from exportTable to produce CNA calls?

Hi. I am afraid not. and the problem is in "similar adjacent windows". It is difficult to decide what is "similar" and what is "different enough".

what is, more precicely, the problem you have? Too many segmnents? Over-fragmentation?

Amjad85 · 01-07-2013, 01:48 AM

Quote:

Originally Posted by stefanoberri

Hi. I am afraid not. and the problem is in "similar adjacent windows". It is difficult to decide what is "similar" and what is "different enough".

what is, more precicely, the problem you have? Too many segmnents? Over-fragmentation?

Thank you for your reply. I am comparing different algorithms for CNA detection and CNAnorm is one of them. All other algorithms make CNA calls (regions of gain and loss) by merging adjacent windows and reporting mean ratio. So, I am not able to compare CNAnorm to other algorithms with this format

stefanoberri · 01-07-2013, 02:01 AM

I am not sure what you mean. "merging adjacent windows and reporting mean ratio" is usually called segmentation and CNAnorm performs segmentation too (using DNAcopy).
In the file resulted from exportTable is in the last column. You will see several adjacent windows have the same segmented and normalised value. If in doubt please refer to dataset and example in the vignette.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
how to check copy number from certain gene	cliff	Bioinformatics	2	07-14-2014 03:52 PM
exome copy number and LOH	rcorbett	Bioinformatics	20	11-26-2013 01:22 AM
Copy number normalization in capture experiments?	anjulka	General	2	01-23-2012 01:07 PM
tools to detect copy number variants	nans_bn	Bioinformatics	2	01-12-2012 05:25 AM
Copy number analysis using 454 data	ps376	Bioinformatics	3	10-08-2011 05:56 AM

02-02-2012, 09:57 AM	#21
christophe Junior Member Location: Phoenix Join Date: May 2010 Posts: 7	Hi Stefano, Thanks for the CNAnorm Package. As PeakPloidy is really computationally time consuming, is there a way to run it as a parallel job (multithreading) using an R package such as "snow" or similar ? The time to compute a 60MB ".tab" file is long, very long :-) Thanks Chris

01-06-2013, 04:47 AM	#23
Amjad85 Junior Member Location: Finland Join Date: Jan 2013 Posts: 3	Hi, Is there a function to merge the segments (similar adjacent windows) in the output file resulted from exportTable to produce CNA calls?

01-07-2013, 02:01 AM	#26
stefanoberri Member Location: Cambridge area, UK Join Date: Jan 2010 Posts: 35	I am not sure what you mean. "merging adjacent windows and reporting mean ratio" is usually called segmentation and CNAnorm performs segmentation too (using DNAcopy). In the file resulted from exportTable is in the last column. You will see several adjacent windows have the same segmented and normalised value. If in doubt please refer to dataset and example in the vignette.