Usage recommendations, confusion about mosdepth example command #30
Comments
|
Hi @SergeWielhouwer, could you please share with us the header of the SNV-VCF file please. You can use the following command: bcftools view --header-only file.vcf.gz > file_vcf_header.txtWe did test it with Clair3 for the most part (we did one test with Deep Variant) it seems that the header is not correctly formatted, but could be something else, lets check that first |
As you already correctly identified, the "-c X" does only get chromsome X, this should not be there. I will fix that ASAP in the README.
There was a typo in the previous version, which was updated in the latest version on GitHub to 100kb. However, we did not push those changes to pip yet, since we considered them to be minor issues. We will push those changes with the next version of Spectre, which is currently under development. If you want the latest version or bug fixes, please clone and download the latest GitHub version. You can install Spectre locally via pip. Regarding the minimum CNV length itself, we do not recommend going lower than 100kb. Reason being, Spectre was designed to detect large CNVs. Additionally, it depends on how high your signal-to-noise ratio is. Either mapping artifacts or just bad samples could introduce FPs. Best, |
Removed "-c X" from running Mosdepth which only resulted in chromosome X being exported
|
Hi, Thanks for letting me know and the quick fix! I have also included the Spectre log with the SNV option enabled (the cmdline "contig" excluded), hope the info is a bit helpful. Kind regards, Serge |
Hi,
I am currently evaluating Spectre for CNV calling on some GIAB HG002 ONT data and I encountered a few issues which I wasn't able to resolve myself.
Using mosdepth I tried to create required regions.bed.gz file for use with Spectre, which worked fine with the following:
mosdepth -t {threads} -x -b 1000 -Q 20 "variants/{wildcards.sample}/spectre/mosdepth/{wildcards.sample}"' "{input.bam}" 2>{log}'
But, I don't understand why the "-c X" parameter is included in your example command. This would lead to only Chromosome X being evaluated on read depth right? Could you explain why this parameter is there?
When I tried to run Spectre itself I only got some CNVs on chrX and not on any autosomes or other contigs (I don't know the known CNVs for HG002 atm, but was surprised that I only got this result). Any idea what's causing this? The logs mention something about expecting trouble as the Minimum CNV length is smaller than 1000000 base pairs.
I am using GCA_000001405.15_GRCh38_no_alt_analysis_set.fna so I assume the metadata and blacklist are still compatible right? The contig names are identical (chr prefix) and it concerns hg38.
When including a snv file for marking LoH regions I get the following error.
Was this tool tested on Clair3 SNV data? Or should the falsely seen "##cmdline" contig be excluded using zgrep -v or similar to be able to use these vcf files?
I am using:
Btw, spectre should also work on other mammalian species right? Are there any ploidy limitations or such?
Best regards,
Serge
The text was updated successfully, but these errors were encountered: