Hydra-genetics snv_indels module

The snv_indels module consists of a number of variant callers for:
Short-read (SR) data, both somatic and germline small variants, that is SNV and indels.
 Long-read (LR) data, both somatic and germline variants, that is SNV and indels. Most of the variant callers for LR data are based on deep-learning models that were trained on WGS data at low coverage.

Some tools are able to handle data from tumor-only (TO) samples, while others require matched normal (MN) samples.

Somatic calls can be made by: - freebayes, mutect2, vardict, deepsomatic, clairs-to.

Germline calls can be produced by: - haplotypecaller, deepvariant, deeptrio, glnexus.
The module also provide tools for:

  • variant decomposition and normalization with VT (applicable to both SR and LR data),

  • aggregation of the results from different callers into an ensemble with bc_bio (only for SR data),

  • sorting and indexing of the variants with bcftools (applicable to both SR and LR data),

  • predict which somatic variants are likely to be mosaicisms with deepmosaic or mosaicforecast (applicable to both SR and LR data),

  • read-based phasing of variants and haplotagging of reads with whatshap (applicable to both SR and LR data).

The module is designed to be used in a snakemake pipeline and can be easily integrated into existing pipelines that follow the logic of hydra-genetics. Certain tools in this module (e.g. deepsomatic, clairs-to) use functions to compile paths to input BAM files. Read more about these functions here.



Callers

Caller Type Comment
freebayes somatic
mutect2 somatic
vardict-java somatic Java version used due to speed
haplotypecaller germline
deepvariant germline
deeptrio germline trio
glnexus germline trio / joint calling
deepsomatic somatic deep-learning based caller for somatic variants derived from deepvariant
deepmosaic somatic predict mosaicisms in TO samples based on the output of deepsomatic or Mutect2
mosaicforecast somatic predict mosaicisms in TO samples based on the output of deepsomatic or Mutect2
clairs-to somatic deep-learning based caller for somatic variants in TO samples

Bioinformatic steps available in the module

Short-read data

Steps

Long-read data

Steps

Module input files

Aligned, duplicate marked, sorted, and indexed .bam files. The .bam files are either split into chromosomes or merged.

  • alignment/picard_mark_duplicates/{sample}_{type}_{chr}.bam
  • alignment/samtools_merge_bam/{sample}_{type}.bam
  • (reference genome reference/{genome}.fa)

Module output files

Small variants in the .vcf- or .bcf-format for respective caller or one ensembled, decomposed, and normalized .vcf file.

Heterogeneity in the VAF-related fields and in the way of computing the VAF are handled by the rule fix_af and by the tool bc_bio in the ensembling step: * snv_indels/bcbio_variation_recall_ensemble/{sample}_{type}.ensembled.vcf.gz * snv_indels/deeptrio/{sample}_{type}/{trio_member}.merged.sorted.vcf.gz * snv_indels/deepvariant/{sample}_{type}.merged.sorted.vcf.gz * snv_indels/freebayes/{sample}_{type}.merged.sorted.vcf.gz * snv_indels/gatk_mutect2/{sample}_{type}.merged.sorted.vcf.gz * snv_indels/glnexus/{sample}_{type}.vcf.gz * snv_indels/haplotypecaller/{sample}_{type}.merged.sorted.vcf.gz * snv_indels/vardict/{sample}_{type}.merged.sorted.vcf.gz


Small variants in the .vcf format that can be decomposed and normalized as well. Possibly handled into an ensemble as well, but this may require to reprocess the VCF fields. Indeed, some callers output for instance VAF as AF and others as VAF, and sometimes the VAF is part of the INFO field, while in other cases it is in the FORMAT field: * snv_indels/deepsomatic_tn/{sample}_{type}.vcf.gz
* snv_indels/deepsomatic_t_only/{sample}_{type}.vcf.gz * snv_indels/deepmosaic/{sample}_{type}/final_predictions.txt * snv_indels/mosaicforecast/{sample}_{type}/all.phasing * snv_indels/mosaicforecast/{sample}_{type}/SNP.predictions * snv_indels/clairs_to/{sample}_{type}.snv-indels.vcf.gz * snv_indels/deepvariant/{sample}_{type}.merged.vcf.gz * snv_indels/deepvariant/{sample}_{type}.merged.g.vcf.gz


Small variants in GVCF format:

  • snv_indels/deeptrio/{sample}_{type}/{trio_member}.g.vcf.gz
  • snv_indels/gatk_mutect2_gvcf/{sample}_{type}.merged.g.vcf.gz


Phased variants and haplotagged reads:

  • snv_indels/whatshap_phase/{sample}_{type}.phased.vcf.gz
  • snv_indels/whatshap_haplotag/{sample}_{type}.haplotagged.bam