SVDSS experiments

Prerequisites

Tools

conda install minimap2 pbsv=2.6.2 sniffles=1.0.12 cutesv=1.0.11 svim=1.4.2 dipcall samtools bcftools pysam biopython numpy pandas seaborn
pip3 install truvari

# We need ngmlr v0.2.8 (master) due to a bug in previous releases (not yet in conda at the time of the experiments)
git clone https://github.com/philres/ngmlr.git
cd ngmlr
git checkout a2a31fb6a63547be29c5868a7747e0c6f6e9e41f
mkdir build ; cd build
cmake ..
make

# pbmm2 and debreak will be loaded by snakemake from envs/ folder due to conflicts

# clone and install SVDSS from https://github.com/Parsoa/SVDSS

Data

In our experimental evaluation we used hg38.

Reference and annotations

# Reference
wget https://hgdownload.cse.ucsc.edu/goldenpath/hg38/bigZips/hg38.fa.gz
gunzip hg38.fa.gz
samtools faidx hg38.fa chr1 chr2 chr3 chr4 chr5 chr6 chr7 chr8 chr9 chr10 chr11 chr12 chr13 chr14 chr15 chr16 chr17 chr18 chr19 chr20 chr21 chr22 chrX chrY > hg38.chroms.fa
sed -i '/^[^>]/ y/BDEFHIJKLMNOPQRSUVWXYZbdefhijklmnopqrsuvwxyz/NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN/' hg38.chroms.fa
samtools faidx hg38.chroms.fa
cut -f1,2 hg38.chroms.fa.fai | sort -k 1 > hg38.chroms.fa.genome

# Dipcall PAR
wget https://raw.githubusercontent.com/lh3/dipcall/master/data/hs38.PAR.bed

# pbsv TRF regions
wget https://raw.githubusercontent.com/PacificBiosciences/pbsv/master/annotations/human_GRCh38_no_alt_analysis_set.trf.bed

Tiers

In our paper we used the Tier 1 from GIAB and then we created an Extended Tier 2, that is everything outside Tier 1. Tier 1 is downloaded from GIAB FTP server and lifted to hg38. Extended Tier 2 (here called Tier23) is computed by complementing Tier 1.

# Get Tier 1 and lift to hg38
wget https://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/data/AshkenazimTrio/analysis/NIST_SVs_Integration_v0.6/HG002_SVs_Tier1_v0.6.bed
sed -e 's/^/chr/' HG002_SVs_Tier1_v0.6.bed > Tier1_v0.6.hg19.bed
wget http://hgdownload.soe.ucsc.edu/admin/exe/linux.x86_64/liftOver
chmod +x liftOver
wget http://hgdownload.soe.ucsc.edu/goldenPath/hg19/liftOver/hg19ToHg38.over.chain.gz
gunzip hg19ToHg38.over.chain.gz
./liftOver Tier1_v0.6.hg19.bed hg19ToHg38.over.chain Tier1_v0.6.hg38.bed Tier1_v0.6.unlifted.bed

# Clean and sort bed to be able to complement it
for c in $(seq 1 22) X Y ; do grep -P "^chr${c}\t" Tier1_v0.6.hg38.bed ; done | sort -k1,1 -k2,2n > Tier1_v0.6.hg38.noalt.bed
bedtools complement -i Tier1_v0.6.hg38.noalt.bed -g hg38.chroms.genome > Tier23_v0.6.hg38.noalt.bed

Truth VCFs

For users convenience, we provide the three VCFs computed with dipcall against hg38 and used as groundtruth in our analysis (see truthsets folder).

config.yaml

The experiments are provided as Snakemake workflows. The file config.yaml is used to tweak snakemake execution:

• name: custom name for the run
• fa: reference genome (i.e., hg38.chroms.fa)
• fq: HiFi sample
• trf: trf annotation
• par: PAR regions
• hap1: first haplotype
• hap2: second haplotype
• tier1: GIAB Tier 1
• tier23: Extended Tier 2
• out: output directory (everything will go here)
• pingpong_bin: path to SVDSS binary
• ngmlr_bin: path to ngmlr (v0.2.8) binary
• threads: number of threads for each rule that requires more threads
• aligners: list of aligners to use
• callers: list of callers to use

HG007

Download corrected HiFi reads and corresponding haplotypes:

• HG007.asm.hap1.p_ctg.fa
• HG007.asm.hap2.p_ctg.fa

Change config.yaml accordingly and run:

# Mappers + callers
snakemake --use-conda [-n] -p -j 16
# Dipcall and truvari
snakemake -s Snakefile.dipvari [-n] -p -j 8

Coverage titration for 5x and 10x

samtools view -b -s 0.3 /path/to/out/dir/pbmm2.bam > pbmm2.5x.bam
samtools fastq pbmm2.5x.bam > pbmm2.5x.fq
# change config.yaml
snakemake --use-conda -p -j 16
snakemake -s Snakefile.dipvari -p -j 8

samtools view -b -s 0.6 /path/to/out/dir/pbmm2.bam > pbmm2.10x.bam
samtools fastq pbmm2.10x.bam > pbmm2.10x.fq
# change config.yaml
snakemake --use-conda -p -j 16
snakemake -s Snakefile.dipvari -p -j 8

To plot results, assuming ${OUT-*} to be the output folders set in config.yaml, run:

python3 plot_coverage.py ${OUT-5x}/pbmm2/results.csv ${OUT-10x}/pbmm2/results.csv ${OUT}/pbmm2/results.csv

Other plots

Assuming ${OUT} to be the output folder set in config.yaml:

# VCF length distribution
python3 plot_vcflen.py ${OUT}

# Upset plot for Ext. Tier 2 region
python3 plot_upset.py ${OUT}/pbmm2/truvari/ tier23conf > hg007.tier23conf.ppexclusive.vcf

# Exclusive calls from SVDSS heterozygosity
python3 check_exclusive.py get hg007.tier23conf.ppexclusive.vcf ${OUT}/pbmm2/truvari/dipcall.tier23conf.vcf > hg007.tier23conf.ppexclusive.dist
python3 check_exclusive.py plot hg007.tier23conf.ppexclusive.dist

HG002

Download corrected HiFi reads and corresponding haplotypes:

• HG002.asm.hap1.p_ctg.fa
• HG002.asm.hap2.p_ctg.fa

Update config.yaml accordingly and run snakemake.

CMRG analysis

To compare the callsets to the CMRG callset:

# Download the CMRG callset
wget https://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/release/AshkenazimTrio/HG002_NA24385_son/CMRG_v1.00/GRCh38/StructuralVariant/HG002_GRCh38_CMRG_SV_v1.00.vcf.gz
wget https://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/release/AshkenazimTrio/HG002_NA24385_son/CMRG_v1.00/GRCh38/StructuralVariant/HG002_GRCh38_CMRG_SV_v1.00.vcf.gz.tbi

# Run truvari against the CMRG callset (assuming ${HG2-OUT} to be the output folder for the HG002 analysis)
for t in pp cutesv pbsv svim sniffles debreak ; do truvari bench -b HG002_GRCh38_CMRG_SV_v1.00.vcf.gz -c ${HG2-OUT}/pbmm2/${t}.vcf.gz -o ${t} -f ~/data/hg38-refanno/hg38.chroms.fa -r 1000 -p 0.00 -s 20 -S 20 ; done

# Plot the CMRG venn and supervenn
python3 medical_analysis.py . # . is the folder with the folders created by truvari in the previous cycle

GIAB vs dipcall analysis

To compare the tools against the GIAB callset, download the hg19 data and run the pipeline.

Download (hg19) reference and annotations:

wget ftp://ftp-trace.ncbi.nih.gov/1000genomes/ftp/technical/reference/phase2_reference_assembly_sequence/hs37d5.fa.gz
gunzip hs37d5.fa.gz
# Extract chromosomes
samtools faidx hs37d5.fa 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X Y > hs37d5.chroms.fa
# Map non-ACGT characters to N:
sed -i '/^[^>]/ y/BDEFHIJKLMNOPQRSUVWXYZbdefhijklmnopqrsuvwxyz/NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN/' hs37d5.chroms.fa

# Get hg19 PAR region from dipcall repository:
wget https://raw.githubusercontent.com/lh3/dipcall/master/data/hs37d5.PAR.bed

Download GIAB Tier 1 and complement it:

# Get GIAB VCF and tiers:
wget https://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/data/AshkenazimTrio/analysis/NIST_SVs_Integration_v0.6/HG002_SVs_Tier1_v0.6.vcf.gz
wget https://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/data/AshkenazimTrio/analysis/NIST_SVs_Integration_v0.6/HG002_SVs_Tier1_v0.6.vcf.gz.tbi
wget https://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/data/AshkenazimTrio/analysis/NIST_SVs_Integration_v0.6/HG002_SVs_Tier1_v0.6.bed
# Here we just complement the Tier 1 to get the Extended Tier 2
bedtools complement -i HG002_SVs_Tier1_v0.6.bed -g hs37d5.chroms.fa.fai > HG002_SVs_Tier23_v0.6.bed

Then, update config.yaml accordingly (see config-hg19.yaml, i.e., everything should point to hg19 and a new output directory) and:

# run everything on the older reference release
snakemake --use-conda -p -j 16
# compare the new results with dipcall (against hg19)
snakemake -s Snakefile.dipvari -p -j 8
# compare the new results with the GIAB callset, GIAB vs dipcall, and dipcall vs GIAB
snakemake -s Snakefile.giabvari -p -j 8

# Heterozygosity plot
python3 plot_hetebars.py {dipcall.clean.vcf.gz} {GIAB.clean.vcf.gz} ${OUT}/pbmm2/

CHM13

Download HiFi reads and corresponding haplotypes.

Update config.yaml accordingly and run snakemake.