No cnv found in teh vcf

[pcr876]

Hi I have know cnvs. I am running spectre for my targeted sequenced data for cancer. I am not getting the output what I expected to see. Also I have trouble understanding the images. Following is the command I used:

spectre CNVCaller \ --coverage spectre/mosd.regions.bed.gz \ --blacklist grch38_blacklist_spectre.bed \ --metadata spectre/metadata.mdr \ --sample-id spectre_101 \ --only-chr chr1,chr2,chr5,chr6,chr7,chr9,chr10,chr15,chr17,chr19,chrX \ --cancer \ --disable-max-coverage \ --output-dir spectre/ \ --reference reference/GCA_000001405.15_GRCh38_no_alt_analysis_set.fasta

Also I have attached the images I have obtained running spectre:

Thank you in advance,

Shreya

[philippesanio]

Hi @pcr876

Thanks for reaching out.

The output of the images should just give you an idea where the CNVs are located and how the signal performs. It sould look something like this:

The pink lines describe the DEL/DUP thresholds, and the black line is the normalized coverage mean * ploidy state of the chromosome.

Based on our output images (chr7 and chr9) it seems there is something off.

• chr7 only has a couple of data points.
• chr9 has almost no data points (small peak 21,970,000-22,010,000 = 40kb)

Questions:

• What version of Spectre are you running?
• What is your bin size from Mosdepth?
• How does your coverage data for chr7 and chr9 look like?
• Do other chromosomes look okay?

When running Spectre in the cancer mode, we highly recommend you using SNV data in the form of a VCF, which can be provided with the --snv. Spectre then calculates the copy number neutral state from the SNV data instead of the coverage. We did that since cancer samples can be highly mutative, which would result in enormous coverage ranges in the genome.

Best, Philippe

[eesiribloom]

please see my question about SNV files with cancer/tumour data #34 . Thanks in advance