SVJedi : Genotyping structural variations with long read data

Note [June 2023]: SVJedi has been replaced by SVJedi-graph, a newer version that is faster and improves the genotyping quality especially for close and overlapping SVs.

Go to https://github.com/SandraLouise/SVJedi-graph

SVJedi is a structural variation (SV) genotyper for long read data. Based on a representation of the different alleles, it estimates the genotype of each variant in a given individual sample based on allele-specific alignment counts. SVJedi takes as input a variant file (VCF), a reference genome (fasta) and a long read file (fasta/fastq) and outputs the initial variant file with an additional column containing genotyping information (VCF).

SVJedi processes deletions, insertions, inversions and translocations.

SVJedi is organized in three main steps:

1. Generate representative allele sequences of a set of SVs given in a vcf file
2. Map reads on previously generated allele sequences using Minimap2
3. Genotype SVs and output a vcf file

Jedi comes from the verb jediñ ['ʒeːdɪ] in Breton, it means calculate.

Requirements

• Python3
• Minimap2 (we recommend using version 2.17-r941)
• NumPy
• Biopython

Usage

python3 svjedi.py -v <set_of_sv.vcf> -r <refgenome.fasta> -i <long_reads.fastq>

Note: Chromosome names in reference.fasta and in set_of_sv.vcf must be the same. Also, the SVTYPE tag must be present in the VCF (SVTYPE=DEL or SVTYPE=INS or SVTYPE=INV or SVTYPE=BND). More details are given in SV representation in VCF.

Installation

git clone https://github.com/llecompte/SVJedi.git

SVJedi is also distributed as a Bioconda package:

conda install -c bioconda svjedi    

Examples

The folder Data/HG002_son includes an example of 20 SVs (10 insertions and 10 deletions) to genotype on a subsample of a real human dataset of the Ashkenazim son HG002.

Example command line:

python3 svjedi.py -v Data/HG002_son/HG002_20SVs_Tier1_v0.6_PASS.vcf -a Data/HG002_son/reference_at_breakpoints.fasta -i Data/HG002_son/PacBio_reads_set.fastq.gz -o Data/HG002_son/genotype_results.vcf

Note: Genotyping results in Data/HG002_son/expected_genotype_results.vcf were obtained using Minimap2 version 2.17-r941.

The folder Data/C_elegans includes an example on 12 SVs (del, ins, inv, bnd) to genotype with a small synthetic read dataset on a subset of the Caenorhabditis elegans genome.

Example command line:

python3 svjedi.py -v Data/C_elegans/test.vcf -r Data/C_elegans/genome.fasta -i Data/C_elegans/simulated-reads.fastq.gz

Parameters

SVJedi two different usages from non aligned reads or from aligned reads (PAF format).

    python3 svjedi.py -v <set_of_sv.vcf> -r <refgenome.fasta> -i <long_reads.fastq>

    python3 svjedi.py -v <set_of_sv.vcf> -a <refallele.fasta> -i <long_reads.fastq>

    python3 svjedi.py -v <set_of_sv.vcf> -p <alignments.paf>

SV representation in VCF

Here are the information needed for SVJedi to genotype the following SV types. All variants must have the CHROM and POS fields defined, with the chromosome names in reference.fasta and in set_of_sv.vcf that must be the same. Then additional information is required according to SV type:

• Deletion

  • Either ALT field is <DEL> or INFO field must contain SVTYPE=DEL
  • INFO field must contain either END=pos (with pos being the end position of the deleted segment) or SVLEN=len (with len being the size of the deletion) tags

• Insertion

  • INFO field must contain SVTYPE=INS
  • ALT field must contain the sequence of the insertion

• Inversion

  • Either ALT field is <INV> or INFO field must contain SVTYPE=INV
  • INFO field must contain END=pos tag, with pos being the second breakpoint position

• Translocation

  • INFO field must contain SVTYPE=BND and CHR2= and END= tags
  • CHR2 name and sequence must be in the reference genome fasta file
  • ALT field must be formated as: t[chr:pos[, t]chr:pos], ]chr:pos]t or [chr:pos[t, with chrand pos indicating the second breakpoint position and brackets directions indicating which parts of the two chromosomes should be joined together

Reference

SVJedi: Genotyping structural variations with long reads. Lecompte L, Peterlongo P, Lavenier D, Lemaitre C. Bioinformatics 2020, 36(17):4568–4575 doi:10.1093/bioinformatics/btaa527 (bioRxiv preprint)

Contact

SVJedi is a Genscale tool developed by Lolita Lecompte. For any bug report or feedback, please use the Issues form of Github.