SV calling and vcf annotation

[leahkemp]

Hi all!

I'm currently looking into somatic structural variant (SV) calling with their intended final destination being loaded into scout.

I'm wondering if you might be able to provide an overview of your general workflow in terms of variant calling and vcf annotation. I'm not looking to get too fancy, I just want to call some structural variants and annotate the vcf's with the fields/annotations scout parses into it's user interface.

Currently I'm calling SV's with Manta (implemented through parabricks for analysis speedups).

Next is the annotation. I've been looking at the user interface of the scout's demo instance (as well as the underlying vcf fles) to see what kind of annotations are present, to try and figure out what to annotate my vcf's with.

From what I can gather from the annotations and the vcf header info, it looks like you're calling SV's with manta and annotating the resultant vcf with VEP and genmod and something with bcftools? The below is a snippet of the vcf header of the demo clinical SV vcf file from which I'm making inferences.

##Software=<ID=genmod,Version=3.5.9,Date="2016-12-08 17:36",CommandLineOptions="processes=4 keyword=Annotation family_file=<open file '/mnt/hds/proj/bioinfo/develop/henrikS/internal_id-miptest/analysis/internal_id-miptest/internal_id-miptest.fam', mode 'r' at 0x7fc018606c00> family_type=cmms reduced_penetrance=<open file '/mnt/hds/proj/bioinfo/develop/references_4.0/cust003-CMMS-red-pen.txt.csv', mode 'r' at 0x7fc018606b70> outfile=<unopened file '/dev/stdout' w> temp_dir=/scratch/717377 variant_file=/scratch/717377/internal_id-miptest_sorted_md_brecal_comb_vep_parsed_SV.selected.vcf logger=<logging.Logger object at 0x7fc018338c50> vep=True">
##VEP=v85 cache=/mnt/hds/proj/bioinfo/components/maintainance/miniconda/envs/mip4.0/ensembl-tools-release-85/cache/homo_sapiens/85_GRCh37 db=. ESP=20141103 HGMD-PUBLIC=20152 polyphen=2.2.2 regbuild=13 dbSNP=144 ClinVar=201507 assembly=GRCh37.p13 gencode=GENCODE 19 sift=sift5.2.2 genebuild=2011-04 COSMIC=71
##bcftools_annotateCommand=annotate --header-lines /mnt/hds/proj/bioinfo/develop/references_4.0/vcfanno_headerLines.v1.0.txt /scratch/717374/internal_id-miptest_sorted_md_brecal_comb_SV_vt_filt_genmod_filter_vcfanno.vcf
##bcftools_annotateVersion=1.3.1+htslib-1.3.1
##bcftools_concatCommand=concat -a /scratch/717374/internal_id-miptest_pmanta.vcf.gz
##bcftools_concatVersion=1.3.1+htslib-1.3.1
##bcftools_viewCommand=view -f PASS /scratch/717374/internal_id-miptest_sorted_md_brecal_comb_SV_vt.vcf
##bcftools_viewVersion=1.3.1+htslib-1.3.1
##cmdline=/mnt/hds/proj/bioinfo/components/maintainance/miniconda/envs/mip4.0/bin/configManta.py --bam /scratch/717373/ADM1059A1_lanes_1_sorted_md_brecal.bam --bam /scratch/717373/ADM1059A2_lanes_1_sorted_md_brecal.bam --bam /scratch/717373/ADM1059A3_lanes_1_sorted_md_brecal.bam --referenceFasta /mnt/hds/proj/bioinfo/develop/references_4.0/Homo_sapiens.GRCh37.d5.fasta --exome --runDir /scratch/717373

Apologies if I've missed some obvious documentation somewhere, I've had a look but didn't find the answers to my questions :)

[northwestwitch]

Hello! I feel like I'm not super up-to-date with the latest specifics of the SV calling. Perhaps @dnil can answer this question better than me?

[dnil]

Heh, let's see. We long ago decided to document all of this as tests, so it would be live tested that everything works while being obvious what is used, but it didn't get enough prio.

We have two local pipelines that produce SV calls for Scout viewing currently, MIP (RD) and Balsamic (somatic). Other collaborators using Scout have similar ones (e.g. https://github.com/Clinical-Genomics-Lund/nextflow_wgs), and a new (inter)national pipe is being developed (https://nf-co.re/raredisease).

They each use a somewhat complementary set of callers ( TIDDIT, CNVnator, Manta, Delly, CNVkit and ASCAT). The calls are then typically merged with SVDB and annotated with VEP and to some extent VCFanno, and ranked with Genmod.

The perhaps most important take home for SV calling is you need to get good local database frequency annotation going. We use SVDB and/or loqusdb.

We have loaded cases with SVs from the Sarek pipe as well on the somatic side, and RD ones from FindSV, though we have not maintained the latter for some time.

That should give you a starting point, but I'd recommend perhaps in particular to consider looking at, forking and/or joining the now in progress https://github.com/nf-core/raredisease pipe development. We haven't really tested this well with scout yet, but we are committed to getting it working anyway, and it should smooth and future proof the connection a bit, compared to brewing your own. Which should still work if you prefer it of course!

[leahkemp]

Wicked! Thank you for the incredibly useful information. I'll be poking around having a closer look at things this week, so I'll contribute my thoughts and possibly ask more questions as they arise 😃