severus for the tumor only mode
qinqian opened this issue · comments
Dear authors,
Thank you for developing such a great software. I am using severus to look at our own data, we only have tumor only sample, but in your pipeline, you mentioned the phased vcf should always come from the normal sample, could we use the tumor sample to generate input for the phased vcf? Also, could you please provide the command used in the following benchmarking section? Thank you!
![image](https://private-user-images.githubusercontent.com/884386/297528497-85504e84-9b13-4b3f-ae23-33f9b8b69138.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjM5MTE2NjgsIm5iZiI6MTcyMzkxMTM2OCwicGF0aCI6Ii84ODQzODYvMjk3NTI4NDk3LTg1NTA0ZTg0LTliMTMtNGIzZi1hZTIzLTMzZjliOGI2OTEzOC5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwODE3JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDgxN1QxNjE2MDhaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT0yZjJjZjRlNWNmZmU1MTY2MTIwNGQ5ZDQyYTNkZTViOGJlOTViOWYzMmZlYjViYzJjN2M3ZTM5YmJmYWFmNWVmJlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.BsXpPM5sPOS_e9gfJl_Poh61oMLsi5kxYl2ORz4RW4o)
Best wishes,
Alvin
Hello,
That is for somatic mode. In tumor-only mode, you can use phased vcf from the tumor.
For the benchmarking, we used Minda (https://github.com/KolmogorovLab/minda), which will be available soon.
Hello @aysegokce,
Thank you so much for your quick reply! This clears our confusion for the tumor-only mode command. I may be wrong, I used to think tumor only model is equivalent to somatic mode, but from your explanation, they are different, do the following command line modes look right for severus?
tumor only: severus --target-bam phased_tumor.bam --out-dir severus_out -t 16 --phasing-vcf phased_tumor.vcf
--vntr-bed ./vntrs/human_GRCh38_no_alt_analysis_set.trf.bed
somatic mode: severus --target-bam phased_tumor.bam --out-dir severus_out -t 16 --phasing-vcf phased_tumor.vcf
--vntr-bed ./vntrs/human_GRCh38_no_alt_analysis_set.trf.bed --only-somatic, is this the mosaic mode as in Sniffles2?
germline mode for either tumor or normal sample:
severus --target-bam phased_tumor.bam --out-dir severus_out -t 16 --phasing-vcf phased_tumor.vcf --vntr-bed ./vntrs/human_GRCh38_no_alt_analysis_set.trf.bed
severus --target-bam phased_normal.bam --out-dir severus_out -t 16 --phasing-vcf phased_normal.vcf --vntr-bed ./vntrs/human_GRCh38_no_alt_analysis_set.trf.bed
tumor-normal pair: severus --target-bam phased_tumor.bam --control-bam phased_normal.bam --out-dir severus_out
-t 16 --phasing-vcf phased_normal.vcf --vntr-bed ./vntrs/human_GRCh38_no_alt_analysis_set.trf.bed
Could you please explain the key difference between somatic mode and tumor only model? Maybe it is worth clarify these modes in more details for the Quick Usage section.
Thank you for sharing the Minda github repo. Looking forward to Minda release soon.
Best,
Alvin
Hello @qinqian ,
Germline mode is when there is no control sample (tumor-only, normal-only), and somatic is for a tumor/normal pair.
In the somatic run, Severus outputs both somatic and germline SVs, and --only-somatic
is to output only somatic SVs, which is more useful in large multi-sample studies.
For tumor only runs:
severus --target-bam phased_tumor.bam --out-dir severus_out -t 16 --phasing-vcf phased_tumor.vcf
--vntr-bed ./vntrs/human_GRCh38_no_alt_analysis_set.trf.bed
For somatic:
severus --target-bam phased_tumor.bam --control-bam phased_normal.bam --out-dir severus_out
-t 16 --phasing-vcf phased_normal.vcf --vntr-bed ./vntrs/human_GRCh38_no_alt_analysis_set.trf.bed
Sorry for the confusion.
Ayse
Thanks @aysegokce a lot for your explanation! Never mind, I learned a lot from your messages.
Another question is about the difference between the two benchmarking sections section 2 and section 3, is the section 2 showing the tumor-only mode? How to define somatic mutation when there is only tumor sample? Are you using gnomad af filtering?
In section 2, we used a multi-platform truth set to compare the performance of SV callers. In section 3, we sequenced cancer cell lines and their matching normals and generated an ensemble list for SV calls from four available somatic SV callers. In both sections, we have tumor and normal samples.
We have single sample calls in section 1. In the tumor-only mode, we don't define somatic mutations. We output complex events which are likely to be somatic.
Ayse
Thanks Ayse a lot! It looks crystal clear now.
Hi Ayse,
I tried tumor-only mode with severus (version 0.1.1) with tumor sample phasing vcf, but there is zero SV in the somatic_SVs/vcf, but there are plenty of SVs in the all_SVs/.vcf, is this normal? Thanks!
Best,
Alvin
Hello Alvin,
Yes, we often get around 20k SVs (germline + somatic). In tumor-only cases, since there are no normal, Severus only outputs all_SVs. Depending on tumor purity, subclonal events (<0.4 vaf or <0.8 hVAF) are more likely to be somatic, whereas homozygous events (GT 1/1) are likely to be germline (if the coverage is preserved in the region).
Ayse
Hi Ayse,
Thank you so much for your explanation!
Best,
Alvin