[BioC] cn.mops: copy number for low-coverage bam files?
Günter Klambauer
klambauer at bioinf.jku.at
Tue Apr 30 22:20:02 CEST 2013
Hello Paul,
Thanks for your interest in cn.mops. It should be possible to identify
CNVs quite reliably at this coverage, if you use the right parameters.
The parameter "WL" in the function "getReadCountsFromBAM" is crucial for
the analysis, because it determines the length of the segments, in which
the reads are counted. If the coverage is low, the segments should be
longer. Which "WL" (window length) did you use?
The second important parameter that controls sensitivity and specificity
is "priorImpact". I suppose you expect more CNVs in your case, therefore
you should lower the parameter. The parameter "minReadCount" is not
crucial for the algorithm, and should be left to default.
Did you use the function "cn.mops" or "referencecn.mops"?? Do you have
two samples (tumor and normal) or do you have more tumor samples and
more normal samples. In the first case you should use "referencecn.mops"
and in the second case "cn.mops".
Please do not hesitate to ask, if you encounter any problems.
Regards,
Günter
On 04/30/2013 06:26 PM, Paul Shannon wrote:
> I'd like to take advantange of the segmentation algorightms in cn.mops, to estimate copy number from low-coverage BAM files. The average coverage in our tumor and normal DNA, is about 0.8 and 0.7 respectively, quite uniformly across the whole genome. The consistency of the reads, and their ratios, allow us to plausibly identify amplifications and deletions, either visually, or by simple calculations. Now it's time for more rigor, for which I turned to cn.mops.
>
> However, it appears that such low-coverage sequencing is apparently not well-suited to cn.mops.
>
> The man page says:
>
> minReadCount: If all samples are below this value the algorithm will, return the prior knowledge.
> This prevents that the algorithm, from being applied to segments with very low coverage. Default=1.
>
> Even if I set minReadCount to 0.2, no useful results are returned:
>
> Individual CNVs:
> GRanges with 2 ranges and 4 metadata columns:
> seqnames ranges strand | sampleName median mean CN
> <Rle> <IRanges> <Rle> | <factor> <numeric> <numeric> <logical>
> [1] chr10 [ 42441001, 42553000] * | F7272LUNG.sorted.bam 0.5849625 0.5767351 <NA>
> [2] chr10 [135478001, 135534746] * | F7272SCLC.sorted.bam 0.2196978 0.5403949 <NA>
>
> Before I give up on using this impressive package, may I ask if there is indeed a way to use it with low-coverage but otherwise high-quality bam files?
>
> Thanks!
>
> - Paul
>
More information about the Bioconductor
mailing list