Cnv Analysis After Segmentations

Question

[Irsan]

Hi there evryone,

We have tumor-normal paired WG-seq data and have successfully estimated copy numbers and segmented them which basically leaves us with a list of coordinates. Now my question is what are best practices after segmentation? We are planning to annotate the segments by downloading transcripts located at the CNVs with the GenomicFeatures package from Bioc and then do a gene set enrichment analysis. Should I be doing some extra quality filtering like filtering my segments for known CNVs, low mappability regions, ...

Anyone wants to share code or best practices?

[Sean Davis]

How many samples? Whole genome? What depth?

[Irsan]

Only 1 tumor-normal pair (pilot study) whole genome, 40x, 100bp paired en reads (insert size 100 bp)

[Chirag Nepal]

Hi Irsan,

Could you please provide a sample (working tutorial) on how you can get the final list of amp/del genes in n number of T/N samples. I used varscan2 to analyze copynumber call and used CBS. I am stuck here at the moment.

Thanks !

[Irsan]

Ask this as a new question or do a google/biostar search

Answer 1

[Sean Davis]

I'd suggest taking a look at a paper like Characterizing complex structural variation in germline and somatic genomes. It sounds like you have done a basic read count based analysis. You may want to look at PEM methods as well as split-read methods to be complete. Since you have a single cancer tumor/normal, you'll want to identify genes that appear to be activated or inactivated by somatic SNVs and small indels and structural rearrangements, not just those with copy number changes. I'd suggest a focus on those techniques since they are likely to be more actionable in the setting of a single sample. As you increase your sample size, you can then apply techniques such as GISTIC to define better regions of shared copy number states. Note GSEA on copy number data from a single sample is not likely to be fruitful since the vast majority of genes in such copy number variable regions are likely passengers and perhaps only half of them are even expressed.

[rse]

Hi Sean, do you know how to apply gistic to ngs data? I have the segmented file. But how to get the number of markers and marker file?